Income Diversification in the Chinese Banking Industry: Challenges and Opportunities [1st ed.] 9789811558894, 9789811558900

Table of contents :
Front Matter ....Pages i-xiv
Introduction (Zhixian Qu)....Pages 1-23
The Rise and Development of Non-traditional Banking Business in China (Zhixian Qu)....Pages 25-52
The Impact of Income Diversification on Chinese Banks: Bank Performance (Zhixian Qu)....Pages 53-88
The Impact of Income Diversification on Chinese Banks: Bank Risk (Zhixian Qu)....Pages 89-117
The Impact of Income Diversification on Chinese Banks: Bank Efficiency (Zhixian Qu)....Pages 119-154
Conclusion (Zhixian Qu)....Pages 155-161
Back Matter ....Pages 163-167

Citation preview

Zhixian Qu

Income Diversification in the Chinese Banking Industry: Challenges and Opportunities

Income Diversification in the Chinese Banking Industry: Challenges and Opportunities

Zhixian Qu

Income Diversification in the Chinese Banking Industry: Challenges and Opportunities

123

Zhixian Qu China Construction Bank Beijing, China

ISBN 978-981-15-5889-4 ISBN 978-981-15-5890-0 https://doi.org/10.1007/978-981-15-5890-0

(eBook)

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

China’s financial industry is currently undergoing a period of rapid transformation: heterogeneous service and technology innovations are continuously emerging, and the process of financial liberalization is urging Chinese financial institutions to develop new operating systems. At the same time, the opening up of China’s financial industry to foreign capital has acted as a strong impetus for reform of the financial system, accelerating the pace of transformation and upgrading of the whole industry. As a result, as the main body of China’s financial market, commercial banks are undergoing profound changes in business philosophy and competitive environment. New strategies and business models are forthcoming, and comprehensive management has become an important strategic choice for the financial industry. Since the implementation of the Financial Services Modernization Act by the U.S. government in 1999, mixed operation has become the mainstream in the international banking market. In China, although the supervision model is based on “separate operations and separate regulations,” the development of financial disintermediation, interest rate liberalization, and new types of Internet finance are gradually blurring the boundaries between various financial sub-markets, and promoting the innovative integration of different types of financial fields. In addition, China’s economic development has entered a “new normal,” as the investment-orientation phase gradually gives way to a new growth phase centered on innovation and efficiency. In the context of macro policy changes, commercial banking, as a “regulator” of the national economy, is bound to usher in a new round of reforms. As income structure gradually shifts from the traditional credit business to guarantees, securities, insurance, financial derivatives, and other intermediate businesses, mixed operation has become a new “blue ocean” for bank strategic transformation. However, we should also note the potential risks associated with the process of diversification. The practice of financial markets in developed countries reveals that the innovative business is usually accompanied by complex financial instruments, the high correlation and leverage characteristics of which are more likely to exacerbate the risk aggregation of the financial system, resulting in black swans, v

vi

Preface

gray rhinos or other unexpected events. The report of the 19th National Congress of the Communist Party of China, which is the outline of China’s development from 2017 to 2022, emphasizes the construction of a modern financial system, and gives priority to “improving the financial regulatory system” and “defusing major risks.” Therefore, as Chinese commercial banks enter new and unfamiliar financial sub-industries, regulators should give close consideration to questions such as how to improve the quality and efficiency of banking businesses, and how to avoid systemic financial crises caused by market and regulatory failures in a context of increasingly large scale of assets and ever more complex business lines. These problems have brought new tasks and challenges to banks’ risk management and regulatory agencies. In my new book, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, I systematically analyse and summarize the exploration and practice of mixed business strategy in the Chinese banking industry. The book begins with a history of the reform of the operational system of China’s financial industry, and an analysis of the motivation and development process of commercial banks’ diversification. Then, focusing on the three dimensions of profitability, risk management, and internal efficiency, the book considers in depth the banks’ mixed business strategy from both theoretical and empirical perspectives. Overall, this book provides thorough and comprehensive explanations of relevant theories, and gives readers a panoramic view of the opportunities, challenges, and corresponding solutions that the banking industry faces in the process of entering a wider and more complex financial environment. I hope that the publication of this book could be of great benefit to financial practitioners and business operators, and to general readers with an interest in this field. Beijing, China

Zhixian Qu

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 The Focus of the Book . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Main Contributions of the Book . . . . . . . . . . . . . . . . . . 1.2.1 Content Improvements . . . . . . . . . . . . . . . . . . . . 1.2.2 Methodology Improvements . . . . . . . . . . . . . . . . 1.3 Income Diversification in the Chinese Banking Industry: An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 Classification of Diversification . . . . . . . . . . . . . 1.3.2 Channels of Income Diversification . . . . . . . . . . 1.3.3 Income Diversification Activities . . . . . . . . . . . . 1.4 Theoretical Basis and Endogeneity Problem . . . . . . . . . . 1.4.1 Theories on Income Diversification . . . . . . . . . . . 1.4.2 Endogeneity Problem . . . . . . . . . . . . . . . . . . . . . 1.5 Categorization of the Chinese Banking Market . . . . . . . . 1.6 Book Map and Chapter Synopses . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

1 2 3 3 5

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

7 7 8 9 10 10 15 16 18 19

... ... ...

25 25 26

...

27

...

28

. . . .

30 32 32 33

2 The Rise and Development of Non-traditional Banking Business in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Banking System Reforms in China . . . . . . . . . . . . . . . . . . . . 2.2.1 First Stage of Reform: Establishment of a “Two-Tier” Banking System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Second Stage of Reform: Instituting the Regulatory Framework of “One Bank and Two Commissions” . . . 2.2.3 Third Stage of Reform: Towards a Market-Oriented Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Changes in the Income Structure of China’s Banks . . . . . . . . 2.3.1 Structural Changes in the Mature Banking Market . . . . 2.3.2 Income Diversification in the Chinese Banking Market

. . . .

. . . .

vii

viii

Contents

2.4 Motivations for Chinese Banks’ Diversification 2.4.1 Constraint-Induced Diversification . . . . 2.4.2 Internal-Bank Motivations . . . . . . . . . . 2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

38 38 44 48 49

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

53 53 56 59 59 61 62 66

....

66

3 The Impact of Income Diversification on Chinese Banks: Bank Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Research Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Variables, Data and Methodology . . . . . . . . . . . . . . . . . . . . 3.3.1 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Data Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . 3.4 Empirical Research and Analysis . . . . . . . . . . . . . . . . . . . . . 3.4.1 Income Diversification and Performance: Whole Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Income Diversification and Performance Across Chinese Banking Groups . . . . . . . . . . . . . . . . . . . . . 3.4.3 Effects of Diversification on Performance by Components of Non-interest Activities . . . . . . . . . 3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

....

71

.... .... ....

78 85 85

4 The Impact of Income Diversification on Chinese Banks: Bank Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Related Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Data Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Empirical Estimation and Results . . . . . . . . . . . . . . . . . . . . 4.4.1 Estimation Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Overall Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 Components of Non-interest Income and Risk . . . . . . 4.4.4 Income Diversification and Different Types of Risk . . 4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

89 89 91 94 94 94 96 100 100 101 106 110 113 114

5 The Impact of Income Diversification on Chinese Banks: Bank Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 5.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

Contents

ix

5.3 Variables, Data and Methodology . . . . . . . . . . . . . . . . . 5.3.1 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Data Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Investigation Strategy and Research Methodology 5.4 Empirical Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Estimating Efficiency Scores for Chinese Banks . 5.4.2 Determination of Diversification Effects on Cost and Profit Efficiency . . . . . . . . . . . . . . . . . . . . . 5.4.3 Discussion of the Results . . . . . . . . . . . . . . . . . . 5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Conclusion . . . . . . . . . . . . . . . . 6.1 Main Findings . . . . . . . . . . 6.2 Implications of the Research 6.3 Limitations and Avenues for References . . . . . . . . . . . . . . . . .

................ ................ ................ Future Research . . . ................

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

125 125 127 127 133 133

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

137 148 149 150

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

155 155 158 160 161

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

List of Figures

Fig. 2.1 Fig. 2.2

Fig. 2.3

Fig. 2.4

Fig. 2.5

Fig. 2.6

Fig. 2.7

Fig. 2.8

Financial regulatory framework in China. Source Author’s creation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Volume changes of interest and non-interest income and the growth rate of non-interest income for the big-five banks from 2008 to 2017. Source Author’s creation according to annual reports of the big-five banks . . . . . . . . . . . . . . . . . . Volume changes of interest and non-interest income and the growth rate of non-interest income for joint-stock banks from 2008 to 2017. Source Author’s creation according to annual reports of 12 joint-stock banks . . . . . . . . . . . . . . . . The income structure of China’s banking industry in 2016. Source Author’s creation according to China Banking Regulatory Commission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Changes in the growth rate of fee-based income and its components for the big-five banks from 2008 to 2017. Source Author’s creation according to annual reports of Big-Five banks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Changes in the growth rate of fee-based income and its components for joint-stock banks from 2008 to 2017. Source Author’s creation according to annual reports of 12 joint-stock banks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Benchmark interest spreads in China from 1991 to 2015. Notes Interest spread = Benchmark loan rate − Benchmark deposit rate. Source Author’s creation according to the People’s Bank of China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ROE and profit after tax in the Chinese banking sector from 2007 to 2017. Source Author’s creation according to China Banking Regulatory Commission . . . . . . . . . . . . . . .

..

29

..

35

..

35

..

36

..

37

..

37

..

39

..

39

xi

xii

Fig. 2.9

Fig. 2.10

Fig. 2.11

Fig. 2.12

Fig. 2.13

Fig. 3.1

Fig. 5.1 Fig. 5.2

List of Figures

The NPL ratio and capital adequacy ratio of average Chinese banks and foreign banks from 2010 to 2017. Source Author’s creation according to China Banking Regulatory Commission and annual report of each foreign banks . . . . . . . . . . . . . . . . . Excess reserves in the Chinese banking sector from 2003 to 2017. (The People’s Bank of China did not report the excess reserves for large and joint stock banks in 2010). Source Author’s creation according to the People’s Bank of China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Growth of non-governmental financing and its structure from 2003 to 2015. Source Author’s creation according to National Bureau of Statistics of China . . . . . . . . . . . . . . . . Capitalisation of the Chinese stock markets. Source Author’s creation according to Shanghai stock exchange and Shenzhen stock exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Non-performing loans by Chinese banks from 2010 to 2016. Source Author’s creation according to China Banking Regulatory Commissions annual reports . . . . . . . . . . . . . . . . . Chinese banks’ non-interest income. Notes Operating income = Net interest income + non-interest income. Source Author’s calculations based on BankScope database. . Yearly mean cost efficiency for the whole sample and three sub-groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yearly mean profit efficiency for the whole sample and three sub-groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

..

43

..

45

..

46

..

46

..

47

..

54

. . 137 . . 137

List of Tables

Table Table Table Table Table

2.1 2.2 2.3 3.1 3.2

Table 3.3 Table 3.4 Table 3.5 Table 3.6 Table 3.7 Table 3.8 Table 3.9 Table 3.10 Table 3.11 Table 3.12 Table 3.13 Table 4.1 Table 4.2 Table 4.3

Types of non-interest activities in early stage . . . . . . . . . . . . Accounting indicators of Bank of China, 2015 and 2016 . . . The first batch of five pilot private banks . . . . . . . . . . . . . . . List of the banks in three categories . . . . . . . . . . . . . . . . . . . Descriptive statistics for Chinese banks from 2005 to 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and profitability for Chinese banks, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results for three components of non-interest income and bank performance for Chinese banks, 2005–2016 . . . . . Descriptive statistics for Chinese G-SIBs, 2005–2016 . . . . . Income diversification and profitability for Chinese G-SIBs, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Descriptive statistics for Chinese D-SIBs, 2005–2016 . . . . . Income diversification and profitability for Chinese D-SIBs, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Descriptive statistics for Chinese N-SIBs, 2005–2016 . . . . . Income diversification and profitability for Chinese N-SIBs, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results for three components of non-interest income and bank performance for G-SIBs, 2005–2016 . . . . . . . . . . Results for three components of non-interest income and bank performance for D-SIBs, 2005–2016 . . . . . . . . . . Results for three components of non-interest income and bank performance for N-SIBs, 2005–2016 . . . . . . . . . . Income diversification and risk for Chinese banks from 2007 to 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and risk for Chinese G-SIBs, D-SIBs and N-SIBs, 2007–2016 . . . . . . . . . . . . . . . . . . . . . Income diversification and risk for 35 Chinese banks, G-SIBs, D-SIBs and N-SIBs from 2007 to 2016 . . . . . . . . .

. . . .

33 42 44 63

..

66

..

67

.. ..

69 71

.. ..

72 74

.. ..

75 76

..

77

..

79

..

81

..

83

. . . .

. . 102 . . 104 . . 108 xiii

xiv

Table 4.4 Table 5.1 Table 5.2 Table 5.3 Table 5.4 Table 5.5 Table 5.6 Table 5.7 Table 5.8

Table 5.9

Table A1

Table A2

Table A3

Table A4

List of Tables

Income diversification and risk for 35 Chinese banks from 2007 to 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Parameter estimates of the cost and profit frontiers . . . . . . . Income diversification and banks’ efficiency for the whole sample, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results for three components of non-interest income and bank efficiency for Chinese banks, 2005–2016 . . . . . . . Income diversification and banks’ efficiency for Chinese G-SIBs, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and banks’ efficiency for Chinese D-SIBs from 2005 to 2015 . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and banks’ efficiency for Chinese N-SIBs, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results for three components of non-interest income and bank efficiency for G-SIBs from 2005 to 2016 . . . . . . . Results for three components of non-interest income and bank efficiency for Chinese D-SIBs from 2005 to 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results for three components of non-interest income and bank efficiency for Chinese NSIBs from 2005 to 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and banks’ performance for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and banks’ performance for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and banks’ efficiency for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Income diversification and banks’ efficiency for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . 111 . . 136 . . 139 . . 140 . . 141 . . 142 . . 143 . . 144

. . 145

. . 146

. . 163

. . 164

. . 165

. . 166

Chapter 1

Introduction

Abstract This chapter begins by giving an overview of the classification and channels of income diversification in the Chinese banking market. Then, it outlines the main motivations of the book, and the research questions to be addressed. After briefly noting the principal findings and potential contributions of this research, the chapter concludes by giving an outline of the composition and organisation of the whole book.

Banks traditionally derive their income through providing deposit and loan services. In the last three decades, however, the global banking industry has experienced a profound change in its business modes. Banks in both mature and developing markets have shifted from their traditional deposit and loan businesses towards nontraditional and fee-based businesses. As a result, the shares of non-interest income in their total revenue have steadily increased. The main factors driving the change are regulatory changes and the changing environment for the banking markets (Allen and Santomero 2001). Other contributing factors include financial liberalization, demand from savers for varied services, and advances in technology (Meslier et al. 2014; Williams 2016). The changes in the business modes of the global banking industry have been on a scale experienced by few industries, and their consequences have been wide ranging. With the shift away from traditional income sources and the expansion into new financial service lines, it is now commonplace for banks to increase the proportion of non-interest income through, for example, underwriting securities, insurance agency, and foreign exchange trading services with high leverage and fewer capital restrictions (Mercieca et al. 2007; Sawada 2013; Lee and Hsieh 2014). The resulting implications are huge and the situation calls for greater research efforts to understand the bank diversification process and its consequences. While some investigations have emerged with regard to mature and developing markets, research on China, currently the largest banking market in the world, has been scarce (Zhou 2014). This book aims to provide a novel perspective on the development of income diversification in the Chinese banking sector. The research focuses on the effects of diversification on the profitability, risk exposure and efficiency of Chinese banks. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0_1

1

2

1 Introduction

1.1 The Focus of the Book In an environment of increasingly tight capital regulation and fierce competition, diversification has become an important avenue for banks in the mature market to develop new income (Amidu and Wolfe 2013). With the adoption of the diversification strategy, these banks have vigorously engaged in a wide range of financial products beyond the traditional ones. These include securities, insurance, trusts, and other financial categories. Over time, their non-traditional business has developed to become an important source of bank revenue, in some cases even exceeding traditional interest income. The process of income diversification has blurred the boundaries between banking and other financial institutions. It has provided banks with the incentive to allocate resources to more profitable activities, which consume less capital but involve high levels of leverage, thus leading to an increase in banks’ exposure to specific and systemic risks and causing inefficiency in resource allocation (Elsas et al. 2010; Brighi and Venturelli 2013). In more detail, Wagner (2010) claims that limited liability incentivizes both bank managers and shareholders to allocate and diversify their portfolio towards correlated assets, and to ignore the risk of joint failures in the banking system brought about by raising bank exposure to common sources of risks. However, such an incentive to managers to over-diversify will in turn harm the stability and allocation efficiency of the wider financial system, because diversification makes institutions more similar to each other, thus exposing them to the same risks and increasing the probability of joint failure (Acharya et al. 2006). Therefore, since the global financial crisis of 2008, the international banking industry has taken steps to implement more prudent management of diversified businesses, and regulators have introduced regulatory changes to improve banking supervision (Delimatsis 2012; Zhu and Chen 2016). China’s banking industry has also experienced the significant rise and development of non-traditional business (Zhou 2014). This has taken place against a background of acceleration of financial reforms, such as interest rate liberalization, banking system reform, and internationalization of the Chinese currency. Increasingly strict capital supervision and financial disintermediation driven by the rapid development of the capital market are forcing Chinese banks to look beyond traditional commercial banking services and to search out new income sources; hence banks are gradually moving away from a single income structure that relies largely on interest income, towards a more diversified income structure. At the same time, Chinese banks have also shifted their focus away from on-balance sheet business to off-balance sheet business, and have started to allocate to their portfolios more highly leveraged and high-yield assets with fewer regulatory capital requirements and activity restrictions (Qu et al. 2017). Chinese banks’ shift to income diversification is still ongoing. While they continue to rely on net interest income, their non-interest income from non-traditional

1.1 The Focus of the Book

3

businesses has started to emerge and is growing. The incipient proportion of noninterest income in total income is relatively small, and the banks have less professional expertise in dealing with the uncertainties inherent in high-leverage activities (Iskandar-Datta and McLaughlin 2007; Barth et al. 2013; Alhassan and Tetteh 2017). Given the transitional nature of China’s financial reform and deregulation of noninterest activities, it is inevitable that the process of diversification and its effects on Chinese banks will have unique characteristics compared to those of banks in other markets. Consequently, it is necessary to investigate the effects generated from income diversification of banks in China, including their profitability, risk exposure and efficiency when pursuing the diversification strategy. In this book, Chinese banks are divided into three sub-groups, according to their level of systemic importance. These are global systemically important banks, national systemically important banks and other banks. Our investigation is focused on the effects of income diversification in three aspects, namely the effects on banks’ profitability, risk level and efficiency. By employing a range of empirical techniques across different sub-samples, we attempt to answer the following research questions: 1. What are the driving forces behind the transformation of income structure in China’s banking industry? Specifically, under the conditions of financial reform and changes in the macroeconomic environment, why do banks increase their involvement in non-interest income? 2. Does the diversification of income structure improve the performance of banks in China? 3. To what extent does income diversification increase banks’ exposure to risk, with regard to bank-specific risks and financial distress? In particular, does there exist a non-linear relationship between Chinese banks’ risk exposure and their diversification levels, and if so what is the threshold and optimal level of noninterest activities for diversified banks to minimize their risk? 4. Several studies have considered banks’ profitability and efficiency together, thus leading to confusion and to a range of inconsistent and contradictory empirical results. This book considers these two aspects separately, asking: Does diversification have the same effect on both profitability and efficiency? Could the process of income diversification increase bank profits, while at the same time causing a reduction in internal efficiency?

1.2 Main Contributions of the Book 1.2.1 Content Improvements This book extends the literature in an important way. The existing literature is limited in its research scope, focusing mainly on mature markets, for example Lepetit et al. (2008), Elsas et al. (2010), Boot and Ratnovski (2012), Nguyen et al. (2016),

4

1 Introduction

and Maudos (2017). Only a small, albeit expanding, body of the literature explores the emerging markets, China in particular. However, compared with the markets in mature and other emerging countries, the Chinese banking industry and regulatory system exhibit a number of unique features. It is these differences that provide the motivation for this study to investigate the income diversification effects in China. In doing so, this book makes several original contributions to the literature. First of all, currently, interest income occupies a large proportion of the income stream of Chinese banks. However, given continual shrinking of the interest margin, combined with the Chinese financial reform and regulatory restriction, in this oligopolistic market banks are driven to find alternative ways to use their market power to make up the loss by binding together their lending and non-interest activities, such as fee and commissions. Therefore, in the Chinese banking market the motivation for income diversification is quite different from that in other markets. Consequently, it is worthwhile to investigate whether or not the composition of the market and the motivation for diversification could lead to different diversification effects on banks’ performance and risks. In the Chinese case in particular, these questions merit a separate and dedicated study. Secondly, the liberalization reform of the Chinese banking system, in particular the relaxation of the access mechanism for capital injection from both foreign and private capital, has created a competitive market atmosphere. Such capital injection supplies technological spillover to the Chinese banking sector and increases the incentive for banks to provide more financial services and to seek the potential benefits from product innovation and diversification. Against this background of changing market ownership structure and the increasing complexity of banking products, it is necessary to investigate the Chinese banking market and to find out how these changes will impact on banks’ performance and risk indicators. Thirdly, compared with other markets, the Chinese banks suffered less negative impact during the global financial crisis. This could be attributed to the Chinese government’s special policy schemes for the banking industry, for example the provision of potential non-performing asset protection, whereby the state-owned asset management companies can help banks to divest non-performing assets; and ‘disguised’ funding through the Huijin Investment Company, which is undertaken by the Ministry of Finance. However, such potential government guarantees not only protect banks from financial crisis, but can also increase agency costs and the too-big-to-fail problem. Consequently, the expansion of banks’ business lines, especially the extension of non-interest business combined with high leverage, carries much higher risk in China than in other markets. Therefore, this book makes a comprehensive analysis of the development of income diversification in the Chinese banking sector in the context of extensive financial reform and regulatory and macroeconomic changes. Furthermore, the channels through which banks can expand their non-interest activities are different in China than in other countries. In the US banking market, a system of bank holding companies has been constructed to facilitate income diversification, while Germany has adopted a universal bank system. However, in the Chinese banking sector these two systems co-exist; that is, state-owned banks and national joint stock banks apply the bank holding companies mechanism, where

1.2 Main Contributions of the Book

5

banks establish or merge with financial intermediates to develop their non-interest activities, while other medium- and small-sized non-systemically important banks adopt the universal banks mechanism, whereby they establish multiple departments within the bank to develop cross-selling business strategies. This mixed development model is a new path that China has chosen in the early stages of diversification. It involves different risk management strategies and the corresponding affordability across different types of banks, hence requires different diversification direction and strategies. In addition, most of the studies conducted with regard to the emerging markets have focused on cross-country datasets (e.g. Gamra and Plihon 2011; Sanya and Wolfe 2011). However, the Chinese banking market has unique features, and should therefore be investigated separately from the emerging market set. Most notably, the scale of the Chinese banking market is now among the largest worldwide: according to the China Banking Regulatory Commission and the World Bank, in 2018 the total assets of China’s banking institutions were 260 trillion CNY (around USD 38 trillion), while the corresponding amount for the Eurozone was USD 31 trillion, and for the USA USD 16 trillion. Clearly, therefore, it is inappropriate to investigate the Chinese market alongside other emerging countries, such as Mexico, Philippines or Thailand, where the magnitude of the market is considerably lower. Finally, unlike banks in other emerging countries, Chinese banks maintain more interconnectedness with the global banking sector. Here, the label ‘systemically important banks’ describes the scale and degree of influence those banks hold in global and domestic financial markets. Currently there are 29 global systemically important banks, of which 13 are in Europe, eight in the US, three in Japan and one in Canada. For the emerging markets, there are only four global systemically important banks, all of which are in China. Therefore, the stability of the Chinese banking sector is of critical importance to the global market. Any credit default in China would be likely to start a chain reaction and cause systemic risk in the global market. It is therefore highly relevant that in recent decades the Chinese banking market has seen an ever increasing growth rate in the expansion of non-interest income; indeed, in 2011, the growth rate of non-interest income reached a high point of 61.9% (CBRC). Such a rapid growth rate is rare in other markets, and brings with it greater uncertainty and other potential factors that could affect the stability of the Chinese market. This makes it particularly important to study the impact of changes in non-interest income in the Chinese market from multiple perspectives, namely those of the banks’ managers, of regulators, and of the stability of the international financial market.

1.2.2 Methodology Improvements In order to avoid the replication of techniques employed in the research of developed markets, this book also applies several improved methodologies to address

6

1 Introduction

bank diversification in the specific context created by China’s unique institutional background and data characteristics. First, in the mature banking market, banks have accumulated sufficient information, techniques and risk management ability to deal with the specific risks generated by the non-interest activities. However, based on the learn-by-doing effect theory, given that Chinese banks only received permission to embrace non-interest activities in 2005, and that the development of non-interest income in the Chinese banking market is very uneven across different banks, it is likely that the relationship between diversification and risk in China is non-linear and follows a dynamic process (Gamra and Plihon 2011), rather than being static in nature as in mature and other emerging markets. This implies that static and linear methods could be inadequate to capture the risk implications of diversification in the Chinese banking market and to estimate the diversification effect on banks’ risk level. Instead, an improved methodology, namely a dynamic threshold model, would be more appropriate to give a clear estimation of the non-linear relation between income diversification and risk, which evolves with the development of banks’ engagement in diversification. As such, a GMM-type threshold model is what is required. Therefore, this book adopts the first-differenced GMM estimator for dynamic threshold panel data model. This research is the first to adopt the above approach to test the learn-by-doing effect in the Chinese banking market. The improved method could help to estimate the potential inverse U-shaped correlation during the diversification process. Compared with traditional static threshold estimation, it has the advantage of giving a dynamic view of the diversification effects on bank risk while avoiding the bias from the quadratic terms used in some previous studies on non-linearity in the banking markets (Bun and Windmeijer 2010; Hsiao and Zhang 2015). The model also overcomes the problems associated with previous GMM-type threshold models (such as Ramírez-Rondán 2015), whereby both regressions and threshold variables have to be exogenous. More importantly, this dynamic threshold estimation could address the endogeneity problem that has been associated with Chinese banking market research, as noted by several previous studies (Acharya et al. 2006; Stiroh and Rumble 2006; Baele et al. 2007). Another important methodology improvement relates to stochastic frontier analysis (SFA), which is usually considered an appropriate tool to investigate the efficiency implications of diversification (Rezitis 2008; Beccalli and Frantz 2009; Cheng 2015). In this book, two issues have been identified and resolved to improve upon the SFA used in the mature market research. First, since the efficiency score generated from SFA falls in the interval [0, 1], it is necessary to model the SFA response properly. Related studies in the mature market generally adopt a static panel data model using the OLS method. However, as the explanatory variable in the regression equation cannot be expected to have a normal distribution, neither can we expect the regression error term to meet the assumption of normal distribution. Consequently, the OLS method often leads to biased and inconsistent parameter estimates (Souza and Gomes 2015). To address these problems, we use a Tobit estimation. In adopting the doubly censored Tobit model developed by Elsas and Florysiak (2015), this study

1.2 Main Contributions of the Book

7

is the first in the field to employ the dynamic estimation of such a model. This estimator addresses the inconsistencies generated from unobserved heterogeneity; furthermore, it allows the addition of a lagged dependent variable, thus providing a dynamic view. Unlike the similar estimation introduced by Loudermilk (2007), it is also applicable for unbalanced panel data, which is particularly important for this research given the restrictions on data collection in the Chinese case. The second issue with regard to SFA is that, given the limited data available for Chinese banking studies, most of the research in this field faces a short panel problem. This leads to an incidental parameters problem, where the variance parameters for short panel data are more likely to be affected under traditional SFA and the fixedeffect SFA model proposed by Greene (2005) and commonly used in the previous literature. Therefore, following Chen et al. (2014), this book solves the problem by adopting within maximum likelihood estimation (WMLE), which relies on the within transformed model using the standard maximum likelihood method. This book represents the first use of WMLE in studying the Chinese case. Finally, the existing literature in this field suffers from a lack of rich analysis. To enrich the analysis and hence ensure the robustness of the research, this book utilizes a large set of banks’ indicators to describe banks’ characteristics. Previous studies, especially those focusing on the Chinese banking industry, utilize only accountingbased measures to assess the diversification effects (e.g. Lepetit et al. 2008; Hsieh et al. 2013). Consequently, the results obtained are not particularly sound or robust. In contrast, this book divides non-interest income into three sub-categories, and investigates efficiency from two perspectives, namely cost and profit efficiency, separately. In addition, this book not only considers accounting data but also employs measures based on economic conceptualization, such as, in the case of risk analysis, both idiosyncratic risk and financial distress. Further, to assess the diversification implications for the management of different types of risk, the research takes into account different aspects of risk, such as credit risk, liquidity risk and interest rate risk. Consequently, this book is able to conduct a more comprehensive analysis of diverse aspects of the effects of income diversification on banks than has been possible in previous research.

1.3 Income Diversification in the Chinese Banking Industry: An Overview 1.3.1 Classification of Diversification For a commercial bank, a diversification strategy means a broadening of income sources, expansion of business scope, and extension of operating activities. Generally speaking, diversification can be classified into three types, namely assets diversification, geographic diversification and income diversification, all of which have

8

1 Introduction

developed in line with technological advances, policy changes and customer demand (DeYoung et al. 2004). Diversification of assets refers to the inclusion of different types of loans within the loan portfolio. Geographic diversification refers to expansion in terms of operational area, where banks set up branches in different regions or countries through establishment or acquisition, and provide financial products and services in local or wider regions to achieve cross-regional or multinational operations (Meslier et al. 2014). Finally, income diversification refers to those activities of banks that are beyond the scope or range of a single financial service product. This type of diversification is manifested by banks’ ability to cross the boundaries of traditional commercial banks in product services and to provide customers with several or all of banking intermediation, securities, insurance, and trusts services (Schmid and Walter 2009). Banks may also update and refine their credit business through financial innovation and may expand into intermediary businesses, securitization and various types of segmentation within each traditional service. With income diversification, banks no longer rely on a single source of income, such as traditional net interest income. Instead, they increase new income through diversified business lines, and diversify the income streams of different businesses through increased non-interest activities. This book focuses mainly on income diversification, and seeks to analyse its effects on Chinese banks.

1.3.2 Channels of Income Diversification At present, there are two channels or business modes of diversification for banks in China. First, large-scale banks tend to adopt the bank holding group mode, which allows them to exploit their advantages in scale, outlets and customers to build a cross-market and diversified financial services platform (Peng and Hu 2005). Banks build up platforms for non-traditional activities through acquisitions, holdings, or the establishment of financial leasing companies, trust companies, fund management companies, insurance companies, and other non-bank financial institutions in both the domestic and overseas markets. Meanwhile, financial companies can enter the banking market and establish financial holding groups. Currently, among the main players in the Chinese banking market the Everbright Group, CITIC Group, Ping An Group, and Shanghai International Group each hold a full license to access all financial services in China, and are the owners of, or majority shareholders in, the China Everbright Bank, China CITIC Bank, Ping An Bank, and Shanghai Pudong Development Bank, respectively. Second, in an environment of strict financial restrictions, many medium and smallsized commercial banks are turning their attention to cooperation with trust, security, fund insurance, and other non-banking financial institutions (Hachem and Song 2015). Under the widely adopted pattern of bank-trust cooperation, the bank sells the issued wealth management products to investors, and the funds raised are passed

1.3 Income Diversification in the Chinese Banking Industry …

9

to the trust company. Then the trust company invests the funds in a company designated by the bank. Since 2009, regulations concerning such cooperation have been tightened, and this has prompted banks to look for cooperation with other financial institutions. First of all, bank-security cooperation is aimed at transferring credit assets (mainly bill assets) to off-balance sheets (Lu et al. 2015; Xu 2017). Here, banks use wealth management funds to purchase securities companies’ asset management plans, which they use to put funds into designated projects to avoid the constraints of the loan-to-deposit ratio control and credit scale imposed by the Chinese regulator. Bank-insurance cooperation in China follows an agency sales model, where commercial banks sell insurance products on behalf of insurance companies, and in return receive commission income (Haifeng 2011).

1.3.3 Income Diversification Activities For banks, income sources can be divided into interest business and non-interest business (Mamun and Hassan 2014). Interest business is business related to the bank’s net interest income and includes traditional deposits and loans businesses. Non-interest income refers to all income other than that from loans and securities business. According to Chinese regulations, the non-interest income of banks consists of five parts: net fee and commission income, investment income, fair value exchange income, exchange gains, and other business income. A large proportion of noninterest income is generated from fee-based activities; that is, it is revenue obtained by charging customers for certain financial services. This includes, among others, monthly service fees for trading accounts, commissions for insurance coverage for homes and businesses, membership fees for the acceptance and use of certain types of credit cards, and income from financial consulting services for individuals and companies. Fee income can be further classified into two main categories: that originating from the traditional business and services of commercial banks, such as cheque and savings account fees, machine usage fees, and fees and commissions for providing loans to customers; and that originating from, and expanding with, non-traditional businesses, such as fees and commissions for investment banking services, trading products, investment products such as stocks, bonds and mutual funds for clients, and service fees for providing wealth management services for customers through affiliated trust company departments. In short, bank diversification in China means that banks diversify their income sources, continually expand their scope of financial services, and increase the proportion of non-interest income in the total revenue through, for example, underwriting securities, securitization, and foreign exchange trading services, with financial innovation and higher financial leverage (Laeven and Levine 2007; Doumpos et al. 2016).

10

1 Introduction

1.4 Theoretical Basis and Endogeneity Problem 1.4.1 Theories on Income Diversification Several theories have been proposed to explain the effect of bank diversification, and whether this brings benefits or discount to banks’ performance, risk and efficiency. According to Boot (2003) and Meng et al. (2017) diversification can be treated as a strategic response to business uncertainty. Hence, most studies in this field are based on the modern portfolio theory and suggest a risk separation effect, thus giving a positive view of the diversification effect on banks’ efficiency. Further, several scholars highlight that diversification can help banks to gain benefits through improved informational advantages (Akhigbe and Stevenson 2010), increased the market power (Palich et al. 2000), the construction of an internal capital market (Pitelis 2007), and economies of scale (Drucker and Puri 2008). However, other researchers claim that there is a diversification discount, whereby the benefits pointed out by portfolio theory might be eliminated by the presence of asymmetric information (Shen and Lee 2006), moral hazard and agency problems (Freixas et al. 2007), rent-seeking behaviours (Datta et al. 2009), high-intensity market competition, and joint failure under the condition of business homogenization (Acharya et al. 2006). Finally, owing to the learn-by-doing effect (Lou 2008), diversification might exhibit a non-linear relationship with banks’ performance.

1.4.1.1

Modern Portfolio Theory

Portfolio theory, the most widely used theory to explain banks’ diversification activities, suggests that increasing the proportion of non-interest income can provide a potential risk reduction. Modern portfolio theory indicates that concentrated revenue streams adversely impact banks’ revenue volatility, and that a strategy of income diversification could generate a coinsurance effect (Lewellen 1971; Tong 2012). As suggested by Mooney and Shim (2015), the coinsurance effect would decrease the volatility of future cash flows for the diversified bank, and make conglomerates less sensitive to the risk taking by a single division. Therefore, banks should improve their stability and disperse idiosyncratic risk through portfolio diversification. Moreover, as argued by Ibragimov et al. (2011), the portfolio theory also suggests that each bank could form a joint mutual market portfolio, whereby each bank would contribute its risky portfolio to the total and receive back its proportional share. In the situation where there was sufficient variety of risk classes, various idiosyncratic risks under individual portfolios would be eliminated. This would result in a more resilient, and more effective, banking system.

1.4 Theoretical Basis and Endogeneity Problem

1.4.1.2

11

Informational Advantages

Diversified banks can obtain superior information from their mixed business lines. According to the theory of financial intermediation, such information advantages represent a further benefit of diversification. More specifically, information is considered an important input factor to impact banks’ efficiency and reduce banks’ specific risk with regard to credit screening and customer relationship (Diamond 1984; Bencivenga and Smith 1991; Saunders and Walter 1994; Elyasiani and Wang 2012). With the increase in informationally intensive assets and financial services, banks gain comparative advantages whereby they can capitalize on client information obtained when they process loans, thus offsetting the excessive credit risks generated from non-interest income, and improving their operational efficiency (Elsas et al. 2010). Where there is integration of lending and non-traditional activities, multiple financial products are sold to similar customers. In this situation each business line could reap benefits from the access to private information and thus reduce the uncertainty associated with the lending relationship (Mercieca et al. 2007). Further, improved information from income diversification within intermediary business could also reduce financial frictions between borrowers and lenders, thus helping banks to overcome asymmetric information (Akhigbe and Stevenson 2010) and gain improved capital financing (Klein and Saidenberg 2010). Pyle (1971) suggested the important interaction between companies’ assets and liabilities, in which asymmetry of information between borrowers and lenders would largely impact on banks’ efficiency characteristics. Diamond (1984) developed a model in which banks could overcome the problem of asymmetric information and improve overall efficiency through mergering with other financial intermediaries; as a result, mature financial intermediation is more likely to minimize the cost of monitoring, which will prove useful to reduce incentive problems between borrowers and lenders.

1.4.1.3

Resource Based View Theory

Based on the transaction cost theory established by Coase (1937) and Williamson (1975), several studies have argued that expansion of the number of types of financial services through the construction of financial conglomerates will help firms to overcome a number of financial problems (e.g. Leff 1978; Hitt et al. 2006). In particular, conglomeration may help member firms to construct an internal capital market and to overcome market imperfections that are prevalent in emerging economies (Khanna and Palepu 2000; Kogut et al. 2002; Wan and Hoskisson 2003; Wan 2005). Therefore, resources integration is an important consideration for banks embarking upon a strategy of diversification. With regard to resources integration, the resource based view (RBV) (Penrose 1959), emphasizes the importance of oligopolistic interaction and interfirm competition (Pitelis 2007). In addition, as argued by Weston (1970), while focused-business

12

1 Introduction

companies can configure resources only through external capital markets, diversified companies are able to increase their effectiveness through access to an internal capital market and by transferring resources to higher profitability objectives, socalled ‘winner-picking’ (Lamont 1997; Stein 1997). Such a winner-picking effect could help banks to reallocate internal resources from less profitable sectors to more effective sectors (Stein 1997).

1.4.1.4

Economies of Scale and Synergistic Effect

According to the economies of scale-based theory proposed by Sirri and Tufano (1995), the expansion of non-interest income can help banks to achieve operational synergies (Rezitis 2008). The attainment of these synergies relies on scale economies (Stiroh 2000; Akhigbe and Stevenson 2010; Sanya and Wolfe 2011). As argued by Drucker and Puri (2008), the expansion of non-interest business is largely based on banks’ infrastructure, hence a mixed business line strategy could help banks to spread fixed costs and managerial overheads over an expanded product mix. Klein and Saidenberg (2010) also suggest that, as several financial intermediaries under diversified banks can separate their contracts to share facilities and production technology, the average cost can be reduced and banks’ profitability and performance can be increased.

1.4.1.5

Asymmetric Information

Because income diversification carries with it the potential problem of asymmetric information between a bank and its pool of borrowers, some scholars have argued that there is likely to be a diversification discount (Sanya and Wolfe 2011). Krishnaswami and Subramaniam (1999) claim that the diversification process greatly increases the asymmetry of information. In addition, Liu and Qi (2003) suggest that diversified firms have insufficient and inadequate channels of information production and transmission to managers, thus reducing the quality of banks’ investment decisions and causing inefficiency and value loss. According to Drucker and Puri (2005), asymmetric information gained through combining lending and underwriting could mean that banks obtain an abnormal return from the securities industry; consequently, banks would have an incentive to underwrite securities of unsound companies and to place them on the financial market without disclosing their private information about the firms (Santos 1998). In addition, due to the very large numbers of customers, banks may not be able to collect sufficient information. As a result, banks may fail to screen out potential bad borrowers, which could compromise banks’ risk prediction and operational strategies, leading to an increase in financial instability (Abdelaziz et al. 2012).

1.4 Theoretical Basis and Endogeneity Problem

1.4.1.6

13

Too-Big-to-Fail and Moral Hazard Problem

The expansion of business scale is also associated with the too-big-to-fail status, which offers a further explanation for the increased risk (Williams 2016). That is, when banks become large enough to be deemed too big to fail, managers have incentives to accentuate the moral hazard, and therefore they may operate the banks in inefficient ways. According to Lin et al. (2012), banks with such moral hazard problems tend to keep large chunks of their resources in less profitable projects, thus causing inefficient allocation of their resources and increased risk. Furthermore, Rezitis (2008) argues that, once a company has achieved a certain scale, it faces inefficient monitoring and supervision across different banking sectors. Especially in emerging countries, banks enjoy invisible guarantees from central banks and governments. Consequently, during financial distresses, banks have incentives to expand high-risk but more profitable projects, which would lead to the accumulation of both specific and systemic risks within the banking system (Hellmann et al. 2000; Kaufman 2014).

1.4.1.7

Structure-Conduct-Performance (SCP) Paradigm

The structure-conduct-performance paradigm is an important theory to explain how the market structure determines the conduct of companies in the market, and then the feedback effects occur such that the company conduct also affects the market structure (Hannan 1991; McWilliams and Smart 1993; Panagiotou 2006; Athanasoglou et al. 2008; Alhassan et al. 2016). In the Chinese banking market, the narrowing spread caused by interest rate liberalization, combined with the ongoing financial disintermediation, increases banks’ willingness to diversify their business in order to broaden their sources of income. In turn, mergers and acquisitions among banks and financial intermediaries, and the consequent departure of banks from traditional business lines, serve to increase business similarities within the financial system and cause income convergence (Ibragimov et al. 2011). Consequently, the diversification process creates competitive pressures amongst financial conglomerates across a wide range of market segments. With regard to the intensely competitive Chinese banking sector, there are two main perspectives, namely the competition-stability view and the competition-fragility view. From the competition-fragility perspective, the more intense competition generated from diversification within the banking industry could make banks less sound. Vives (2011) claims that such a relationship can be explained by the effects through two channels. First, increased interbank competition could increase banks’ frangibility by exacerbating the coordination problem between depositors and the bank. Second, increased competition would change the risk-taking behaviour of the banks. In a more competitive environment with more pressures on profits, there would be increased incentive for the bank to take on more excessive risk on either side of the bank’s balance sheet, resulting in greater fragility. As suggested by Amidu and Wolfe (2013), this can be assumed as an inverse response; that is, the increased competition generated from a diversification strategy could then stimulate a higher level of

14

1 Introduction

diversification, further increasing competition within the banking sector, while at the same time causing the banks to adopt more radical strategies to diversify. Individual banks would thus bear on more insolvency risks, and there would be an increased likelihood of failure.

1.4.1.8

Business Homogenization and Joint Failure

Barry et al. (2011) suggest that bank managers are likely to choose to diversify the banks’ income sources in the expectation that this will separate risk, in order to limit idiosyncratic risk. Hence, managers have incentives to diversify the company beyond the optimal level. However, in doing so they cause harm to the wider financial system, because diversification leads to business homogenization, making institutions more similar to each other and exposing them to same risks, which can lead to joint failure and so are more exposed to the systemic risk (Acharya et al. 2006). Wagner (2010) proposes a contagion model, which detects the conflicts between bank managers’ individual insolvency risk and the systemic risk. The study claims that limited liability incentivizes both bank managers and shareholders to allocate and diversify their portfolio towards correlated assets, and to ignore the risk of joint failures in the banking system brought about by raising bank exposure to common sources of risk. However, such an incentive to managers to over-diversify will in turn harm the wider financial system, because diversification makes institutions more similar to each other, thus exposing them to the same risks and increasing the probability of joint failure (Acharya et al. 2006). Consequently, it creates a fragile financial system, where once a shock or bankruptcy hits an individual bank, the effects would immediately spread and ultimately bring down the whole financial system (Ibragimov et al. 2011). For this reason, diversification might not be conducive to resilience of the banking system, and from a social perspective it might be suboptimal and inefficient (Freixas et al. 2007).

1.4.1.9

Learn-by-Doing Effects

Lee (1996) constructed a learn-by-doing model to investigate the investment and lending decision. According to the model, banks’ behaviour and investment portfolio should be improved through the accumulation of information and employee proficiency. That is, in the early stages, poor information could exist in equilibrium with low investment, leading to an underdevelopment trap; however, this trap would be overcome once the banks had accumulated sufficient learning through experience; that is, ‘learning-by-doing’. Nowadays, the learn-by-doing effects are hardly considered relevant for mature markets, as banks in those markets have already acquired sufficient information and ability to overcome the potential risks and instabilities inherent in non-interest activities. However, given that Chinese banks have only recently received permission to embrace non-interest activities, and are just beginning their diversification process,

1.4 Theoretical Basis and Endogeneity Problem

15

the learn-by-doing theory still have value for research in the Chinese banking sector. As suggested by Lou (2008), the learn-by-doing effect plays a big role in that market, and leads to the possibility of an inverted U-shaped relationship between diversification and banks’ risk level. By international standards, most Chinese banks are in the early stages of income diversification, where without enough information and professionalism the accumulation of high-leveraged non-interest activities will make those banks less stable. At the same time, some banks may have crossed the diversification threshold and already be enjoying the benefits from diversity. That is, in the early stages of expansion of non-interest activities banks suffer risk-enhancement, but if they continue with the process and pass a certain threshold level, then, assuming that they have gained rich experience and professionalism, and that a sound regulatory system is in place, they can reap a risk discount.

1.4.2 Endogeneity Problem More recent studies on the diversification effect in the banking industry place strong emphasis on the two-way relationship between income diversification and banks’ other specific characteristics. Such a mutual causality will lead to deterioration of the econometric model, due to the endogeneity bias (Sanya and Wolfe 2011; Köhler 2014). A number of studies have found that high-risk banks are more likely to enter into riskier and high-leveraged non-interest activities, and to propose a more radical diversification strategy (e.g. Lang and Stulz 1994; Acharya et al. 2006). Similarly, Stiroh (2004) state that when facing high volatility of earnings, banks are more likely to expand their businesses scale and to implement merger and acquisition with high-leveraged financial sectors, behaviour that is driven by a risk-taking preference. In addition, with regard to the relationship between performance and diversification, those companies that face constraints upon their business growth and profitability, especially within a highly competitive market, will tend to choose a diversified strategy in order to look for new drivers of profitable growth (Christensen and Montgomery 1981). Furthermore, some market-based factors, such as market competition, or government policy, will have influences on banks’ diversification level and performance and, simultaneously, on risk indicators. For example, intensive market competition narrows banks’ net interest margin, hence greatly shrinks the profitability and performance of the bank’s core business, while at the same time it pushes banks to accelerate the process of diversification. All of the factors mentioned above will lead to a potential endogeneity problem, as will the omitted management strategy variables (Gurbuz et al. 2013) and the sensitivity of bank risk level to macroeconomic shocks (Berger et al. 2000). Therefore, prior studies on the diversification effect in the banking industry have placed important emphasis on this issue (Sanya and Wolfe 2011; Köhler 2014). Therefore, in the empirical sections, this book adopts the generalized method of moments (GMM)

16

1 Introduction

model, which is particularly well-suited to overcoming the inconsistency caused by endogeneity.

1.5 Categorization of the Chinese Banking Market Since the 2008 global financial crisis, bank capital requirements have been tightened and new resolution regimes enabling the orderly failure of banks are being implemented. Particular attention has been given to systemically important banks. In November 2011, the Basel Committee on Banking Supervision (BCBS) finalized its methodology to identify such systemically important banks and the regulatory approach to reduce the economic impact of their default. To identify the global systemically important banks, the BCBS determined five categories according to the many dimensions of systemic importance: size, interconnectedness, substitutability, complexity and the cross-jurisdictional activity of a bank. This led to the identification of 29 large banks as global systemically important banks (G-SIBs); among these, four are Chinese commercial banks. Based on the BCBS categorization, in 2014 the China Banking Regulatory Commission (CBRC) announced three categories of banks in China, and implemented different levels of financial restrictions over them. According to the Guidelines for the Disclosure of Global Systemic Importance Indicators of Commercial Banks, commercial banks with total assets below 1.6 trillion Yuan are defined as non-systemically important (N-SIBs), and should have to satisfy only the basic capital requirement ratios, where the tier one capital ratio and capital adequacy ratio requirements are 5% and 8%, respectively. According to these guidelines, 27 banks fall into the category of non-systemically important banks. Commercial banks with total assets over 1.6 trillion Yuan should be defined as systemically important. According to the Guidelines and the list published as a result of Basel III, these can be categorized further as global systemically important banks (G-SIBs) and domestic systemically important banks (D-SIBs). A total of 13 Chinese commercial banks are classified as systemically important. Of the 13 banks deemed systemically important, four are G-SIBs. Known as ‘the big four’, these G-SIBs, namely the Industrial and Commercial Bank of China (ICBC), Agricultural Bank of China (ABC), Bank of China (BOC) and China Construction Bank (CCB), must possess additional capital that meets or exceeds the uniform requirement of the Basel Committee. According to a China Banking Regulatory Commission Annual Report (2016), owing to the high concentration of the Chinese banking industry, the total assets of these four banks have reached seventy-five billion CNY, accounting for 67.9% of the total assets of the banking industry. In order to maintain the stability of the domestic banking industry, in January 2014 the China Banking Regulatory Commission (CBRC) issued the Guideline for the Disclosure of the Evaluation Index for D-SIBs. This gave a list of D-SIBs and required qualifying commercial banks to disclose their evaluation index. According to the CBRC, nine national banks fall into the category of D-SIBs. Overall, the

1.5 Categorization of the Chinese Banking Market

17

capital requirement for systemically important banks is stricter than for N-SIBs. More specifically, D-SIBs must maintain a minimum 8% core tier one capital and 11.5% capital adequacy ratio. For G-SIBs, the China Banking Regulatory Commission further extended the restrictions to require an additional 1% of risk-weighted assets, which should be satisfied by core tier one capital ratio. In order to test the diversification effect on banks’ specific characteristics and stability this book adopts all thirteen systemically important banks, both global and domestic, along with twenty-seven non-systemically important banks. Together these forty banks account for 79% of the total assets of the Chinese banking industry. As explained above, it is necessary to divide the Chinese banking sector according to the categories G-, D-, and N-SIBs. However, the number of banks in each group is relatively small, especially in the G-SIB category, which includes only four institutions. This could lead to bias, and undermine the validity of the estimation results (Blundell and Bond 1998; Blundell et al. 2001). The main concern with regard to small sample bias is that the group of G-SIBs, which is particularly important for the understanding of Chinese banks’ specific characteristics and stability, includes only four banks, and the data for regression cover only 12 years. However, it is useful and valuable to consider this group as an independent category, as reflected in other studies investigating the Chinese banking market, such as Berger et al. (2010), Chang et al. (2012) and Yin et al. (2013). This is because there exist large differences between G-SIBs and other types of bank in terms of the business scale, total assets and other characteristics. For instance, accroding to the Bankscope database, the Bank of China (BOC), which ranks fourth among the G-SIBs and in the Chinese banking market as a whole, possesses total assets of 3,213 billion USD, and has 311,133 employees. In stark contrast, the Bank of Communications (BOCOM), which ranks first among D-SIBs and fifth in the Chinese banking market, holds total assets of only 1,407 billion USD, less than half that of the BOC, and the number of employees (88,605) is only around a quarter of the number employed by the BOC. Similar differences exist between D- and N-SIBs. For example, all D-SIBs operate nationwide, while all N-SIBs are regional banks. These factors lead to significant differences in the scale of operations, profitability, and the degree of diversity among the three bank groups. A further point that should be mentioned with regard to the small sample size is that, in the Chinese banking sector, the vast majority of banks are unlisted city and rural banks. The data for these banks are often opaque and undisclosed, while only a small number of banks make available the information required by a study such as this one. As a result, the small sample size problem is prevalent in research in the Chinese banking sector: for example, Zhou and Wang (2008) use 12 commercial banks, while Shen et al. (2010) include 14 banks in their sample. In the situation of small sample size, the most popular method, OLS, can lead to bias. According to Soto (2009), in this case GMM is more reliable and efficient than OLS, as the GMM estimator is not hindered even when the small sample size means that it is not possible to completely exploit the linear moment conditions. Therefore, in the empirical part of this book in Chaps. 3 and 4, we adopt GMM estimation in order to strengthen the reliability and efficiency of estimation.

18

1 Introduction

In addition, under the traditional stochastic frontier analysis proposed by Greene (2005) to evaluate banks’ efficiency level, the short panel in the Chinese banking market could lead to an incidental parameters problem and affect the accuracy of estimation. Therefore, in the analysis in Chap. 5, we investigate the efficiency implications of banks’ income diversification by implementing the stochastic frontier analysis with the within maximum likelihood estimation (WMLE) proposed by Chen et al. (2014), in which the first-difference data transformation eliminates nuisance parameters, thus solving the incidental parameters problem and making the estimation of the efficiency scores unbiased. Moreover, in order to verify the potential effect generated from the small sample size, rather than separating the sample into three sub-groups and regressing each of them separately, we conduct the robustness tests for the main empirical results by applying dummy variables to the three groups, as shown in the Appendix for Chaps. 3 and 5. This alternative analysis allows the retention of more samples and information in the estimation, thus avoiding the small sample size problem. However, in Chap. 4, the threshold model does not allow the addition of an interaction term in both the lower and higher regimes, and it is not possible to get more than one threshold value at the same time. Consequently, there is no robustness test in that chapter.

1.6 Book Map and Chapter Synopses This book comprises six chapters. Following this introductory chapter, the rest of the book is structured as follows. Chapter 2 provides a general introduction to the Chinese banking sector and the rise and development of income diversification of Chinese banks. The chapter begins by describing the recent financial reforms and regulatory changes in China. Then, it explores the driving forces behind banks’ engagement in non-interest activities, and gives an overview of their mixed business lines and the incomes thereof. This is the starting point for analysing business diversification in the Chinese banking sector. Chapter 3 analyses the relation between income diversification and banks’ performance. In order to estimate the diversification effects on the performance of different bank groups, three sub-groups are investigated with regard to return and profitability. Chapter 4 explores the diversification-risk nexus in Chinese banks. It applies tests which reveal the existence of a non-linear relation between banks’ idiosyncratic risk, financial distress and diversification level. This chapter also documents that the diversification effect differs among different non-interest activities. Chapter 5 employs advanced stochastic frontier analysis to estimate efficiency score, and dynamic Tobit model with DPF estimator to provide new evidence on whether bank diversification is beneficial to the efficiency score of Chinese systemically and non-systemically important banks. Chapter 6 provides an overall summary of the book. In this final chapter, research findings from earlier chapters are brought together to give an integral picture of

1.6 Book Map and Chapter Synopses

19

the effects of income diversification on Chinese banks. Limitations of the present research and possible avenues for future research are suggested.

References Abdelaziz H, Hichem D, Wafa K (2012) Universal banking and credit risk: evidence from Tunisia. Int J Econ Financ Issues 2(4):496–504 Acharya V, Hasan I, Saunders A (2006) Should banks be diversified? evidence from individual bank loan portfolios. J Bus 79(3):1355–1412 Akhigbe A, Stevenson BA (2010) Profit efficiency in US BHCs: effects of increasing non-traditional revenue sources. Quart Rev Econ Financ 50(2):132–140 Alhassan AL, Tetteh ML, Brobbey FO (2016) Market power, efficiency and bank profitability: evidence from Ghana. Econ Chang Restruct 49(1):71–93 Alhassan AL, Tetteh ML (2017) Non-interest income and bank efficiency in Ghana: a two-stage DEA bootstrapping approach. J Afr Bus 18:124–142 Allen F, Santomero AM (2001) What do financial intermediaries do? J Bank Financ 25(2):271–294 Amidu M, Wolfe S (2013) Does bank competition and diversification lead to greater stability? evidence from emerging markets. Rev Dev Financ 3(3):152–166 Athanasoglou PP, Brissimis SN, Delis MD (2008) Bank-specific, industry-specific and macroeconomic determinants of bank profitability. J Int Financ Mark Inst Money 18(2):121–136 Baele L, De Jonghe O, Vander Vennet R (2007) Does the stock market value bank diversification? J Bank Financ 31(7):1999–2023 Barry TA, Lepetit L, Tarazi A (2011) Ownership structure and risk in publicly held and privately owned banks. J Bank Financ 35(5):1327–1340 Barth JR, Lin C, Ma Y, Seade J, Song FM (2013) Do bank regulation, supervision and monitoring enhance or impede bank efficiency? J Bank Financ 37:2879–2892 Beccalli E, Frantz P (2009) M&A operations and performance in banking. J Financ Serv Res 36:203–220 Bencivenga VR, Smith BD (1991) Financial intermediation and endogenous growth. Rev Econ Stud 58:195–209 Berger AN, Hasan I, Zhou M (2010) The effects of focus versus diversification on bank performance: evidence from Chinese banks. J Bank Financ 34(7):1417–1435 Blundell R, Bond S (1998) Initial conditions and moment restrictions in dynamic panel data models. J Econom 87(1):115–143 Blundell R, Bond S, Windmeijer F (2001) Estimation in dynamic panel data models: improving on the performance of the standard GMM estimator. In: Nonstationary panels, panel cointegration, and dynamic panels. advances in econometrics. JAI Press, Amsterdam Boot AW (2003) Consolidation and strategic positioning in banking with implications for Europe. Working Paper Boot AW, Ratnovski L (2012) Post-crisis banking regulation: evolution of economic thinking as it happened on Vox. CEPR Press, London Brighi P, Venturelli V (2013) How income diversification, size and capital ratio affect bhc’s performance?. University of Modena and Reggio Emilia. Working paper Bun MJ, Windmeijer F (2010) The weak instrument problem of the system GMM estimator in dynamic panel data models. Econom J 13(1):95–126 China Banking Regulatory Commission Annual Report (2016):28–60 Chang TP, Hu JL, Chou RY, Sun L (2012) The sources of bank productivity growth in China during 2002–2009: a disaggregation view. J Bank Financ 36(7):1997–2006 Chen YY, Schmidt P, Wang HJ (2014) Consistent estimation of the fixed effects stochastic frontier model. J Econom 181:65–76

20

1 Introduction

Cheng MY (2015) Non-interest and bank’s efficiency: evidence from Chinese commercial banks. Mod Econ Sci 37(4):49–126 Christensen HK, Montgomery CA (1981) Corporate economic performance: diversification strategy versus market structure. Strateg Manag J 2(4):327–343 Coase RH (1937) The nature of the firm. Economica 4(16):386–405 Datta S, D’Mello R, Iskandar-Datta M (2009) Executive compensation and internal capital market efficiency. Financ Intermed 18(2):242–258 Delimatsis P (2012) Financial innovation and prudential regulation: the new Basel III rules. J World Trade 46:1309–1325 DeYoung R, Hunter WC, Udell GF (2004) The past, present, and probable future for community banks. J Financ Serv Res 25(2–3):85–133 Diamond DW (1984) Financial intermediation and delegated monitoring. Rev Econ Stud 51(3):393– 414 Doumpos M, Gaganis C, Pasiouras F (2016) Bank diversification and overall financial strength: international evidence. Financ Mark Inst Instrum 25(3):169–213 Drucker S, Puri M (2005) On the benefits of concurrent lending and underwriting. J Financ 60(6):2763–2799 Drucker S, Puri M (2008) On loan sales, loan contracting, and lending relationships. Rev Financ Stud 22(7):2835–2872 Elsas R, Florysiak D (2015) Dynamic capital structure adjustment and the impact of fractional dependent variables. J Financ Quant Anal 50(5):1105–1133 Elsas R, Hackethal A, Holzhäuser M (2010) The anatomy of bank diversification. J Bank Financ 34(6):1274–1287 Elyasiani E, Wang Y (2012) Bank holding company diversification and production efficiency. Appl Financ Econ 22:1409–1428 Freixas X, Lóránth G, Morrison AD (2007) Regulating financial conglomerates. J Financ Intermed 16(4):479–514 Gamra SB, Plihon D (2011) Revenue diversification in emerging market banks: implications for financial performance. Unpublished manuscript Greene W (2005) Fixed and random effects in stochastic frontier models. J Product Anal 23(1):7–32 Gurbuz AO, Yanik S, Ayturk Y (2013) Income diversification and bank performance: evidence from Turkish banking sector. J BRSA Bank Financ Mark 7(1):9–29 Hachem K, Song ZM (2015) The rise of China’s shadow banking system. Unpublished manuscript Haifeng G (2011) Study on restructuring risk management mechanism of China small-medium enterprises’ finance guarantee. South China Financ 4:3–20 Hannan TH (1991) Foundations of the structure-conduct-performance paradigm in banking. J Money, Credit Bank 23(1):68–84 Hellmann TF, Murdock KC, Stiglitz JE (2000) Liberalization, moral hazard in banking, and prudential regulation: are capital requirements enough? Am Econ Rev 90(1):147–165 Hitt MA, Tihanyi L, Miller T, Connelly B (2006) International diversification: Antecedents, outcomes, and moderators. J Manag 32(6):831–867 Hsiao C, Zhang J (2015) IV, GMM or likelihood approach to estimate dynamic panel models when either N or T or both are large. J Econom 187(1):312–322 Hsieh MF, Chen PF, Lee CC, Yang SJ (2013) How does diversification impact bank stability? The role of globalization, regulations, and governance environments. Asia-Pac J Financ Stud 42(5):813–844 Ibragimov R, Jaffee D, Walden J (2011) Diversification disasters. J Financ Econ 99(2):333–348 Iskandar-Datta M, McLaughlin R (2007) Global diversification: evidence from corporate operating performance. Corp Ownersh Control 4(4):228–242 Kaufman GG (2014) Too big to fail in banking: what does it mean? J Financ Stab 13:214–223 Khanna T, Palepu K (2000) Is group affiliation profitable in emerging markets? an analysis of diversified Indian business groups. J Financ 55(2):867–891

References

21

Klein PG, Saidenberg MR (2010) Organizational structure and the diversification discount: evidence from commercial banking. J Ind Econ 58:127–155 Kogut B, Walker G, Anand J (2002) Agency and institutions: national divergences in diversification behavior. Organ Sci 13(2):162–178 Köhler M (2014) Does non-interest income make banks more risky? retail-versus investmentoriented banks. Rev Financ Econ 23(4):182–193 Krishnaswami S, Subramaniam V (1999) Information asymmetry, valuation, and the corporate spin-off decision. J Financ Econ 53(1):73–112 Laeven L, Levine R (2007) Is there a diversification discount in financial conglomerates? J Financ Econ 85(2):331–367 Lamont O (1997) Cash flow and investment: evidence from internal capital markets. J Financ 52(1):83–109 Lang HP, Stulz RM (1994) Tobin’s q, diversification and firm performance. J Polit Econ 102:1248– 1280 Lee CC, Hsieh MF (2014) Bank reforms, foreign ownership, and financial stability. J Int Money Financ 40:204–224 Lee J (1996) Financial development by learning. J Dev Econ 50(1):147–164 Leff NH (1978) Industrial organization and entrepreneurship in the developing countries: the economic groups. Econ Dev Cult Chang 26(4):661–675 Lepetit L, Nys E, Rous P, Tarazi A (2008) Bank income structure and risk: an empirical analysis of European banks. J Bank Financ 32:1452–1467 Lewellen WG (1971) A pure financial rationale for the conglomerate merger. J Financ 26(2):521– 537 Lin JR, Chung H, Hsieh MH, Wu S (2012) The determinants of interest margins and their effect on bank diversification: evidence from Asian banks. J Financ Stab 8(2):96–106 Liu Q, Qi R (2003) Information production, corporate investment and diversification discount. Working paper Lou YC (2008) A study of commercial banking’s performance. Stud Int Financ 6:4–11 (in Chinese) Loudermilk MS (2007) Estimation of fractional dependent variables in dynamic panel data models with an application to firm dividend policy. J Bus Econ Stat 25(4):462–472 Lu Y, Guo H, Kao EH, Fung HG (2015) Shadow banking and firm financing in China. Int Rev Econ Financ 36:40–53 Mamun A, Hassan MK (2014) What explains the lack of monetary policy influence on bank holding companies? Rev Financ Econ 23(4):227–235 Maudos J (2017) Income structure, profitability and risk in the European banking sector: the impact of the crisis. Res Int Bus Financ 39:85–101 McWilliams A, Smart DL (1993) Efficiency structure-conduct-performance: implications for strategy research and practice. J Manag 19(1):63–78 Meng X, Cavoli T, Deng X (2017) Determinants of income diversification: evidence from Chinese banks. Appl Econ 1–18 Mercieca S, Schaeck K, Wolfe S (2007) Small European banks: benefits from diversification? J Bank Financ 31(7):1975–1998 Meslier C, Morgan DP, Samolyk K, Tarazi A (2014) The benefits of intrastate and interstate geographic diversification in banking. Unpublished manuscript Mooney T, Shim H (2015) Does financial synergy provide a rationale for conglomerate mergers? Asia-Pac J Financ Stud 44(4):537–586 Nguyen M, Perera S, Skully M (2016) Bank market power, ownership, regional presence and revenue diversification: Evidence from Africa. Emerg Mark Rev 27:36–62 Palich LE, Cardinal LB, Miller CC (2000) Curvilinearity in the diversification–performance linkage: an examination of over three decades of research. Strateg Manag J 21(2):155–174 Panagiotou G (2006) The impact of managerial cognitions on the structure-conduct-performance (SCP) paradigm: a strategic group perspective. Manag Decis 44(3):423–441

22

1 Introduction

Peng SJ, Hu JX (2005) Three choices of Chinese financial industry in the background of industry and finance combination. J Zhengzhou Inst Aeronaut Ind Manag 3:11–35 (in Chinese) Penrose E (1959) The theory of the growth of the firm. Wiley, New York Pitelis CN (2007) A behavioral resource-based view of the firm: the synergy of Cyert and March (1963) and Penrose (1959). Organ Sci 18(3):478–490 Pyle DH (1971) On the theory of financial intermediation. J Financ 26(3):737–747 Qu HJ, Gong T, Chi YP (2017) Capital supervision pressure, income diversification and the stability of listed banks. The Financial Services Forum. Working paper Ramírez-Rondán N (2015) Maximum likelihood estimation of dynamic panel threshold models. In: Peruvian economic association working paper No. 2015-32 Rezitis AN (2008) Efficiency and productivity effects of bank mergers: evidence from the Greek banking industry. Econ Model 25(2):236–254 Santos JA (1998) Commercial banks in the securities business: a review. J Financ Serv Res 14(1):35– 60 Sanya S, Wolfe S (2011) Can banks in emerging economies benefit from revenue diversification? J Financ Serv Res 40(1–2):79–101 Saunders A, Walter I (1994) Universal banking in the United States: what could we gain? what could we lose?. Oxford University Press, Oxford Sawada M (2013) How does the stock market value bank diversification? empirical evidence from Japanese banks. Pac-Basin Financ J 25:40–61 Schmid MM, Walter I (2009) Do financial conglomerates create or destroy economic value? J Financ Intermed 18(2):193–216 Shen CH, Lee CC (2006) Same financial development yet different economic growth–why? J Money, Credit, Bank 38(7):1907–1944 Shen D, Cai F, Zhang L, Zhao L (2010) An empirical study on the motivation of mergers and acquisitions in Chinese banking sector. J Chongqing Technol Bus Univ (Soc Sci Ed) 23(06):42–49 (in Chinese) Sirri E, Tufano P (1995) The economics of pooling. Harvard Business School Press, New York Soto M (2009) System GMM estimation with a small sample Souza GDS, Gomes EG (2015) Management of agricultural research centers in Brazil: a DEA application using a dynamic GMM approach. Europ J Oper Res 240(3):819–824 Stein JC (1997) Internal capital markets and the competition for corporate resources. Journal Financ 52(1):111–133 Stiroh KJ (2000) How did bank holding companies prosper in the 1990s? J Bank Financ 24(11):1703–1745 Stiroh KJ (2004) Diversification in banking: is noninterest income the answer? J Money Credit Bank 36(5):853–882 Stiroh KJ, Rumble A (2006) The dark side of diversification: the case of US financial holding companies. J Bank Financ 30(8):2131–2161 Tong Z (2012) Coinsurance effect and bank lines of credit. J Bank Financ 36(6):1592–1603 Vives X (2011) Competition policy in banking. Oxford Review of Economic Policy 27(3):479–497 Wagner W (2010) Diversification at financial institutions and systemic crises. J Financ Intermed 19(3):373–386 Wan WP (2005) Country resource environments, firm capabilities, and corporate diversification strategies. J Manag Stud 42(1):161–182 Wan WP, Hoskisson RE (2003) Home country environments, corporate diversification strategies, and firm performance. Acad Manag J 46(1):27–45 Weston JF (1970) The nature and significance of conglomerate firms. St. John Law Rev 44:66–80 Williams B (2016) The impact of non-interest income on bank risk in Australia. J Bank Financ 73:16–37 Williamson OE (1975) Markets and hierarchies. New York Free Press, US Xu Z (2017) Does the expansion of interbank business diminish the liquidity of banks in China? Chin Stud 6(01):12–23

References

23

Yin H, Yang J, Mehran J (2013) An empirical study of bank efficiency in China after WTO accession. Glob Financ J 24(2):153–170 Zhou HW, Wang J (2008) A review of the fluctuation of non-interest income of commercial banks of our country from the perspective of portfolio theory. Econ Surv 4:044 (in Chinese) Zhou K (2014) The effect of income diversification on bank risk: evidence from China. Emerg Mark Financ Trade 50(sup3):201–213 (in Chinese) Zhu C, Chen L (2016) An empirical study on the capital buffer of rural commercial banks in China. J Financ Econ 4(3):97–102

Chapter 2

The Rise and Development of Non-traditional Banking Business in China

Abstract This chapter provides the overall picture of the historical process and current status of income diversification in the Chinese banking sector. It begins by outlining the major stages of banking reforms in China. Then, it describes the development of the non-traditional business by Chinese banks. Finally, the driving forces behind Chinese banks’ shift to income diversification are discussed.

The steady growth of non-traditional business has been an important development of China’s banking industry in the recent decades. This takes place amid gradual unfolding of financial reforms in the country including interest rate liberalization, marketization of banking business, and RMB internationalization. Improved banking supervision and financial disintermediation have strengthened the banks’ shift to non-traditional activities. In consequence, Chinese commercial banks increasingly look beyond traditional services and to search out new income sources, leading to a shift from a single income structure that relies largely on interest income, to a more diversified income structure. The first chapter has briefly introduced the channels and main composition of income diversification, and given an overall picture of the organization of this book. In this chapter will describe in detail the major stages of reform of the Chinese banking system, provide an overview of both the historical development and current status of non-interest income, and explain the main causes thereof.

2.1 Introduction In the recent decades, the Chinese financial sector has experienced rapid growth. By the end of 2017, the country’s banking system had become the world’s biggest in terms of bank assets. According to China Banking Regulatory Commission, by 2017, the total assets of China’s banking institutions reached USD 37 trillion, while the corresponding amount for the Eurozone was USD 31 trillion, and USD 16 trillion for the USA. On top of the rapid growth of bank size, great changes have also taken © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0_2

25

26

2 The Rise and Development of Non-traditional Banking Business …

place in banks’ functions, banks’ management, and the regulatory system (Laeven and Levine 2007). Internationally, it is an established trend that commercial banks, especially largesized ones, are actively seeking new income streams and no longer depend on interest income alone (Stiroh and Rumble 2006). In the mature banking markets, the proportion of non-interest income has steadily increased, causing banks to shift their focus away from traditional intermediary functions such as taking deposits, issuing loans, and providing intermediary services, and instead to take on a diversified income structure by providing an extensive range of services. For China, while the general economy has continued to develop, financial reforms are also well under way, which include interest rate liberalization, RMB internationalization and exchange rate system reform (Deng and Luo 2014; Liao and Tapsoba 2014). Such reforming moves have been accelerating, and placing pressure on banks to pursue a strategy to diversify their income sources. Moreover, improved capital regulation and the financial disintermediation due to rapid development of the stock market in the country have further induced commercial banks to end their reliance on the expansion of loan business as the sole source of their profit growth, cementing the trend for shifting to a more diversified income structure. Previous studies lack a comprehensive coverage of the development of income diversification in the Chinese banking market. This chapter intends to fill that gap, by providing an overview of the growth of non-traditional banking business in China including a summary of the special nature of the Chinese regulatory framework and explanations for the internal and external reasons that motivate banks to expand their non-interest income. The Chinese banking industry has undergone profound changes over the past two decades, in terms of changes to the regulatory system, ownership type, market competition and the admission of foreign and private capital. In order to ensure a better understanding of the following chapters it is necessary to have a comprehensive knowledge of this background and development path. Most importantly, during the process of China’s banking reform and policy changes, there have been several significant shifts in the income structure of the banking industry. These changes are very helpful to understand the motivations for Chinese banks’ diversification, and will be referred to frequently in the subsequent chapters.

2.2 Banking System Reforms in China Before 1978, China operated an economic and financial system based on central planning principles. The configuration of the economy followed the Soviet command economy model, whereby economic development was regulated by a strict central planning system, including a centralized state banking system (Yao and Wu 2010). From 1978, the Chinese regulatory authorities implemented several waves

2.2 Banking System Reforms in China

27

of financial reform in the banking sector, with the aim of transforming a policydriven and monopolistic banking system into a modern and market-oriented one (García-Herrero et al. 2009; Jiang et al. 2013). The Chinese central bank, the People’s Bank of China (PBOC), established in 1949, was the only substantial bank in the Chinese banking system to serve the nation’s centrally planned economy (Berger et al. 2009; Zhang et al. 2013). It functioned both as a governmental department implementing decisions made by the State Council, with responsibility for issuing currency and allocating government investment funds, and as a financial institution offering financial services including savings, loans, and the settling of accounts. With over 15000 sub-branches, the PBOC controlled currency in circulation, managed foreign exchange reserves, set the interest rate and took all deposits from, and extended loans to, the public and commercial businesses.

2.2.1 First Stage of Reform: Establishment of a “Two-Tier” Banking System In 1979 China launched the programme of “reform and opening-up”, aiming to satisfy the demand emanating from the more market economic activities emerging from the transition from the centrally planned economy to a market oriented one (Boyreau-Debray and Wei 2005). As part of the transition, the first reforms in the Chinese banking market were aimed at establishing a “two-tier” banking system (Bonin 1999). From 1979 to 1984, three specialized banks, i.e. the Bank of China (BOC), Agricultural Bank of China (ABC) and China Construction Bank (CCB) were spanned off from the PBOC or the Ministry of Finance. In addition, another substantial bank, the Industrial and Commercial Bank of China (ICBC), was established. They are to carry out the general functions of commercial banks though still under state controls. By 1984, these four state-owned banks were well established as commercial banks for different specialities. On top of this, the PBOC was made the central bank of China, performing as the authorities of credit control, currency issuance and monetary policy (Cai et al. 2019). This separation of the commercial banks and monetary authorities resulted in a “two-tier” banking system which designated the status of PBOC as the central bank and specified the operating rules for other banks under which expansion of credit was constrained by their deposits (Lin and Zhang 2009). It also established an institutional framework that bestows the central bank a leading role and the big-four state banks form the backbone of the Chinese banking system (He et al. 2017). To further promote competition and improve the scope of financial services, by 1992, eight small- and medium-sized shareholding commercial banks were allowed to set up. Of them, six are nationwide banks and two are provincial commercial banks and they offer universal bank services to households and firms, mainly in the

28

2 The Rise and Development of Non-traditional Banking Business …

cities. At the same time, in addition to a variety of trust and investment companies, many city credit cooperatives were established at both the central and local level. By the end of 1992, 4800 city credit cooperatives with assets of 187.8 billion RMB emerged. These developments were accompanied by a corresponding increase in the total banking assets, from 151.6 billion RMB to 1140.1 billion RMB for the bank sector as a whole, where the share of non-banks in the total bank assets rose from 9.34 to 14.03% (Almanac of China’s Banking and Finance 1993).

2.2.2 Second Stage of Reform: Instituting the Regulatory Framework of “One Bank and Two Commissions” The second stage of Chinese financial reforms was featured by the intensity of bank supervision. The main aims of the new reforms were to improve banks’ ability to manage financial risk, and to create a competitive and modern banking market. The central bank’s regulatory roles are re-focused to concentrate on bank supervision, and the two newly established commissions would take over the other regulatory functions. In 1993, the government started to change the regulatory focuses of the central bank. Between 1993 and 1997 the newly established China Insurance Regulatory Commission (CIRC) and China Securities Regulatory Commission (CSRC) gradually took over some of the PBOC’s regulatory functions with regard to the insurance and securities businesses. In 2003, the banking regulatory function was transferred from the central bank to the China Banking Regulatory Commission (CBRC), thereby streamlining and transforming the PBOC into a typical central bank charged with formulating and implementing monetary policy and maintaining the price stability (Zhang et al. 2016). While the new system has helped reduce excessive concentration of regulatory powers, the rise of non-interest activities in the Chinese banking sector poses new challenges. The expansion of the non-traditional business makes the boundaries among insurance, banking, securities and financing leasing companies gradually blurred. Because commercial banks and their businesses are regulated by both the CIRC and CBRC, the overlapping led to regulatory confusions and inefficiency. In order to improve the regulatory efficacy and reduce costs for, in 2018 the CBRC and CIRC were combined into the China Banking and Insurance Regulatory Commission (CBIRC), a single commission overseeing both sectors, under the direct control of the State Council. The function of drafting relevant laws and regulations for the banking and insurance industries, which was the responsibility assumed by the CIRC and CBRC, has been taken over by the central bank, the PBOC (Wang 2018). In consequence, these changes led to the emergence of a regulatory framework in China known as “one bank and two commissions”, which is graphed in Fig. 2.1.

Joint Shareholder Commercial Banks Foreign Banks

City Commerc ial Banks

China Banking and Insurance Regulatory Commission

Fig. 2.1 Financial regulatory framework in China. Source Author’s creation

State-owned Banks and Policy Banks

People’s Bank of China

Monetary policy

The State Council

National People’s Congress

Rural Credit Cooperativ es

Non-banks Financial Institutions

China Securities Regulatory Commission

Financial regulation

2.2 Banking System Reforms in China 29

30

2 The Rise and Development of Non-traditional Banking Business …

2.2.3 Third Stage of Reform: Towards a Market-Oriented Economy 2.2.3.1

Reform After the Asian Financial Crisis

Alerted by the Asian financial crisis, the Chinese government has taken measures to address the financial risks to the Chinese financial system since the late 1990s and injected substantial amount of capital to stabilise banking market (Jiang et al. 2013). In 1998, the government issued government bonds to the value of 270 billion RMB to provide a capital injection to increase the capital adequacy ratio of stateowned banks. In addition, four financial asset management companies (AMCs) were established, namely China Cinda Asset Management (CCAM), China Orient Asset Management (COAM), China Great Wall Asset Management (CGWAM) and China HuaRong Asset Management corporate (CHAM), each of which belongs to one of the big-four banks. These AMCs are tasked to taking over distressed bank assets, and employing modern techniques in the recovery of those assets. In the year from 1999 to 2000, the AMCs acquired over 1.4 trillion RMB non-performing loans from the big-four banks, while in the year from 2003 to 2004 they acquired another 1 trillion (Wang 2003). Between 1999 and 2004, the four AMCs also signed debt-toequity swap agreements with over 1100 state-owned enterprises. In detail, the four asset management companies used funds borrowed from the central bank to buy companies’ debt, and then helped the companies to transfer debt to company shares. The funds raised from the sale of those companies’ shares to the public could then be used to repay the debt owed to the central bank. These actions alleviated the potential default risk of state-owned banks and stabilized the Chinese financial system to some extent. Similar to state-owned banks, city credit cooperatives have also accumulated a lot of bad debt and non-performing loans (Hsiao et al. 2015). As credit cooperatives are under direct control of local governments, the central bank could not effectively influence their assets management and loan decisions. Therefore, in 1995, the State Council issued the Notice Concerning the Creation of City Cooperative Banks, under the terms of which the urban credit cooperatives would be merged into new urban commercial banks, and thus brought into the banking regulatory framework. By the end of 2002, over 2000 urban cooperatives were transformed into 111 city commercial banks (Hamid and Tenev 2008). They provide financial services to local citizens, while operating under strict financial supervisions by the central bank.

2.2.3.2

Shareholding System Reform

In 2003, the Chinese government started the Shareholding System Reform, which would transform Chinese commercial banks into shareholding corporations (Bin 2007). This reform was aimed at optimizing banks’ ownership structure to improve

2.2 Banking System Reforms in China

31

financial transparency and operational efficiency (Lin and Zhang 2009; Yao et al. 2008). In early 2004, the State Council utilized foreign exchange reserves to the value of 45 billion USD to inject capital into the COB and CBC for use in financial reorganization and supplementary financing. In 2005, the state-owned Huijin Investment Company injected 15 billion USD into the ICBC. Following further assets restructuring conducted with the assistance of the Minister of Finance, in 2005 the CBC was listed for the first time on the Hong Kong stock exchange (Berger et al. 2009). The following year, the ICBC and BOC were listed on the Shanghai and Hong Kong stock exchanges. Through IPO, state-owned banks could gather funds, increase liquidity and also diversify their ownership (Jia 2009). To date, a total of 38 Chinese banks have been listed on either the overseas or domestic stock exchanges. The change to banks’ ownership structure brought about by IPO means that banks must be responsible to, and ensure they have enough profit and be better at monitoring the risks management for, their shareholders (Boubakri et al. 2005; Clarke et al. 2005). Thus, it stimulates banks to balance their income structure and increase their income stream. At the same time, income diversification could also diversify banks’ portfolios, which could stabilize the income stream and guarantee payments for shareholders. More importantly, under the shareholding system reform, banks also opened up to strategic investors and foreign capital (Tsai et al. 2014). Several banks, such as the Bank of Beijing, have been officially authorised as foreign capital holding banks. The entry of foreign capital has made the business model and income structure of Chinese commercial banks closer to those of foreign banks, as they become more and more efficient and diversified (Berger et al. 2009; Luo and Yao 2010; Luo et al. 2017). In summary, after more than 30 years of gradual financial and banking reforms, China’s banking industry has undergone profound development and significantly improved financial stability, efficiency improvement and economic growth (Hasan et al. 2009; Fang and Jiang 2014; Peng et al. 2014; Wang et al. 2014; Lin et al. 2015), where assets have increased from 304.8 billion RMB in 1978 to 232.25 trillion RMB in 2016. The Chinese banking sector has been transformed from a centrally planned system with one bank that functioned as both policy maker and commercial bank, into a complex two-tier banking system under a regulatory framework with “one bank and two commissions” and with a variety of financial agencies including state-owned banks, private banks, policy banks and other non-bank financial institutions. According to a CBRC annual report, at the end of 2016, China’s banking sector consisted of 3 policy banks, 5 large commercial banks, 12 joint stock commercial banks, 134 city commercial banks, 8 private banks, 1114 rural commercial banks, 40 rural cooperative banks, 1,1125 rural credit cooperatives and 1 postal savings bank. As a result of these developments, combined with the divestiture of policy business from the PBOC and the restructuring of banks’ ownership, the Chinese banking market has been gradually changed from political-oriented to market-oriented, with associated improvements in competitiveness, profitability and diversification level.

32

2 The Rise and Development of Non-traditional Banking Business …

2.3 Changes in the Income Structure of China’s Banks 2.3.1 Structural Changes in the Mature Banking Market Diversification is an important trend in the development of the mature banking market. As such, most banks in the mature market can provide customers with a wide range of financial products and services, including banking, securities, insurance, trusts, and other financial categories. Banks in that market have vigorously pursued a mixed businesses strategy and their non-interest business has developed rapidly, becoming a major source of bank revenue, even exceeding traditional interest income. This fast-moving process of diversification has been driven by several factors. In 1984, the French Banking Act was promulgated, allowing France’s commercial banks to engage in all banking-related financial business and to begin to transition to universal banks. In 1986, the UK authorities announced the Financial Services Act, allowing commercial banks to provide comprehensive financial services including securities and other businesses. The diversification process in the US market began with the Financial Services Modernization Act of 1999, which allowed financial holding companies to operate a variety of financial businesses through the establishment of subsidiaries. The banking markets in Germany and Switzerland have consistently implemented a mixed operation system. The universal banks in those countries operate without any business scope restrictions among financial sectors, and can offer a range of financial services such as commercial and investment banking. The gradual relaxation of financial and businesses regulation in a mixed operation or universal banking system has provided sufficient space for commercial banks to develop their non-interest business and diversify their income streams. Furthermore, the Basel Accord requirements for capital adequacy has forced commercial banks to increase their liquidity, and at the same time to pay attention to the ratio of assets with different risk-weights. In order to meet the capital adequacy ratio requirements, banks shift from on-balance sheet business to off-balance sheet ones and to allocate to their portfolios more high-yield assets, such as investments, venture capital, securitized products and guarantees (Lozano-Vivas and Pasiouras 2014). Therefore, the international strengthening of bank capital requirements tends to induce commercial banks to shift in the income structure from interest-based to non-interest-based income. The rapid development of electronic information technology has also provided favourable technical conditions and a platform for integration of banking and other financial services (Siregar et al. 2017). Technology innovations such as Big Data and Blockchain have made it possible for commercial banks to innovate in financial products and tools, thus facilitating the provision of a more extensive range of financial products and services, the development of e-banking and other online financial platforms, and a consequent expansion of income sources.

2.3 Changes in the Income Structure of China’s Banks

33

2.3.2 Income Diversification in the Chinese Banking Market The development of non-interest income in the Chinese banking sector has evolved through three stages:

2.3.2.1

From 1980 to 1992: Chaotic Development of Business Diversification

Non-interest income includes revenues from commercial banking, investment banking, insurance, asset management, and financial infrastructure services (clearance, settlement, payments, custody, etc.). In the late 1960s, banks in mature economies started the transition from focusing on traditional lending business to a ‘cross-selling’ operational pattern. In 1979, trust and leasing business launched in the Bank of China, and become the first type of non-interest business in Chinese market. During the next few years, commissions for stock issuance and consulting businesses also launched. With the further deregulation of the industry, and in particular the United States’ Financial Services Modernization Act of 1999, commercial banks are allowed to engage in securities underwriting/brokerage, insurance, and other high-leverage areas such as venture capital. Table 2.1 shows the non-interest activities expansion in the early stage of Chinese banking market. Table 2.1 Types of non-interest activities in early stage

Non-interest product

Issuance date

Financial institution

Trust business

1979

Bank of China

Business guarantee

1980

China Construction Bank

Agency for foreign currency trading

1982

Bank of China

Forward exchange transaction

1985

Bank of China

Investment advisory services

1987

China Construction Bank

Interest rates and currency swaps

1988

China CITIC Bank

Revolving underwriting facility

1988

Bank of China

Note issuance facility

1988

Bank of China

Forward rate agreement

1988

Bank of China

Source Author’s creation according to China Statistical Yearbook

34

2 The Rise and Development of Non-traditional Banking Business …

In 1980, the State Council issued the Interim Rules on Promoting Economic Coalition, which allowed the commercial banks to engage in trust business. Subsequently, Chinese commercial banks also obtained permission to enter the securities and insurance businesses. In 1987, the Bank of Communications became the first bank to engage in the comprehensive management of banking, securities and insurance businesses. During the 1980s, banks were keen to develop universal and multi-functional banks, such that the boundaries between banks and other financial institutions became blurred (Lo et al. 2016). With the rapid expansion of non-interest activities and insufficient regulation, management of non-interest business in China’s banking industry were chaotic and the risk management thereof was inadequate.

2.3.2.2

From 1992 to 2000: Restricting Diversification

Partly to address the problems occurred in the banks’ shift to non-interest business, the State Council in 1993 promulgated the Decision on Reform of the Financial System. The document stipulated that commercial banks should be decoupled from the insurance, trust and securities businesses and keep only interest activities in their portfolios. In 1995, the Law of the People’s Republic of China on Commercial Banks further forbade merger activities between banks and financial institutions, thus restricting the expansion of non-interest and brokerage business. With increasingly strict restrictions on mixing banks’ business lines, the proportion of non-interest income shrank accordingly.

2.3.2.3

From 2000: Gradual Reviving of Income Diversification

With the gradual establishment of a comprehensive regulatory system, in 2001 the Chinese authorities turned their attention to income diversification in the banking sector. In that year, the PBOC introduced the Provisional Regulations Governing Commercial Banks’ Intermediary Business, which expanded the business scope for commercial banks and relaxed restrictions on financial derivatives and agency business, including trading in securities, insurance and government bonds. With this groundwork in place, in 2005 the Chinese financial authorities started a pilot program allowing cooperation among financial institutions, thus increasing the complexity and diversification of banks’ income streams. In 2008, the China Banking Regulatory Commission (CBRC) issued the Guidance for Cooperation between the Operations of Banks and Trust Companies, which provided the legal basis to reduce restrictions on bank activity and opened the way for cooperation between non-traditional banking businesses and commercial bank operations. Subsequently, the Securities Regulatory Commission and Insurance Regulatory Commission issued several notices to grant permission for commercial banks to engage in securities and insurance business. Owing to this expansion of non-interest activities, the total volume of non-interest income of the Chinese banking sector grew from 3.57 trillion CNY in 2006 to 25.40 trillion CNY in 2016, a more than six-fold increase over one decade. However,

2.3 Changes in the Income Structure of China’s Banks

35

Million CNY

interest income remains the main source of banks’ total income, and non-interest activities occupy only a moderate proportion of overall revenue. More specifically, according to the CBRC, non-interest income accounts for 23.8% of total operating income of the Chinese banking industry (CBRC Annual Report 2016). Figures 2.2 and 2.3 illustrate the volume and proportion of interest and non-interest income over operating income for the ‘Big-Five’ state-owned banks, and 12 national joint-stock banks, respectively. We see that, for both banking groups, which together constitute the main body of the Chinese banking sector, the proportion of non-interest income has been rising 400000 350000 300000 250000 200000 150000 100000 50000 0

35% 30% 25% 20% 15% 10% 5% 0% 2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Interest and Dividend Income Non-Interest Income Non-interest Income / Operating Income

Million CNY

Fig. 2.2 Volume changes of interest and non-interest income and the growth rate of non-interest income for the big-five banks from 2008 to 2017. Source Author’s creation according to annual reports of the big-five banks

80000 70000 60000 50000 40000 30000 20000 10000 0

40% 35% 30% 25% 20% 15% 10% 5% 0% 2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Interest and Dividend Income Non-Interest Income Proportion of Non-interest Income over Operating Income

Fig. 2.3 Volume changes of interest and non-interest income and the growth rate of non-interest income for joint-stock banks from 2008 to 2017. Source Author’s creation according to annual reports of 12 joint-stock banks

36

2 The Rise and Development of Non-traditional Banking Business …

Fig. 2.4 The income structure of China’s banking industry in 2016. Source Author’s creation according to China Banking Regulatory Commission

1.90 %

1.10 %

6.00 % 17.60 %

73.40 %

Interest Income Exchange Income Investment Income

Fee and Commissions Other Income

steadily year by year. In the case of the joint-stock banks, during the early period the amount of non-interest income was relatively small in terms of both volume and proportion. However, over the past decade, joint-stock banks have paid increasing attention to the development of non-interest income. This has led to rapid growth in the average share of non-interest income, which now exceeds that of the Big-Five state-owned banks. However, to date, the composition of non-interest income is relatively simple, with fees and commissions representing the main sources of non-interest business. As shown in Fig. 2.4, according to the CBRC (2016) fee-based income occupies 17.6% of all income in the Chinese banking sector, followed by investment income (6%), exchange income (1.9%) and other income (1.1%). From the above, it can be seen that fees and commissions are the main sources of the non-interest income in the Chinese banking sector. This fee-based income can be roughly divided into six components: bank card fees; personal wealth management fees; custodian and other fiduciary service fees; settlement, clearing business and cash management fees; investment banking and consultancy fees; and other feebased income. Figures 2.5 and 2.6 show the growth of fee-based income and the volume of each business under fee-based activities, according to figures published in the annual reports of the Big-Five state-owned banks and 12 joint-stock banks, respectively. We see that, as with overall non-interest income, the volume of fee-based income has undergone a significant increase. Specifically, the fee income of the Big-Five banks increased from approximately 4.63 billion USD in 2008 to 14.65 billion USD in 2017, while for the joint-stock banks, fee-based income increased from 0.4 billion USD in 2008 to 5.24 billion USD in 2017. However, we also observe that, with the exception of a rapid increase in the amount of fee income for joint-stock banks during the period from 2008 to 2011, the growth rate of fee-based business in the Big-Five and joint-stock banks has fallen into a continuous decline in recent years. Even in 2017, the average growth rate of the

Million CNY

2.3 Changes in the Income Structure of China’s Banks

37

200000

50%

100000

0%

-50%

0 2008

2009

2010 2011 2012 2013 Other Fee-based Income

2014

2015

2016

2017

Investment Banking and Consultancy Fees Settlement, Clearing Business and Cash Management Fees Personal wealth management, custodian and other fiduciary service fees Bank Card Fees

Fig. 2.5 Changes in the growth rate of fee-based income and its components for the big-five banks from 2008 to 2017. Source Author’s creation according to annual reports of Big-Five banks

Million CNY

40000

100% 50%

20000 0% 0

-50% 2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Other Fee-based Income Investment Banking and Consultancy Fees Settlement, Clearing Business and Cash Management Fees Personal wealth management, custodian and other fiduciary service fees Bank Card Fees

Fig. 2.6 Changes in the growth rate of fee-based income and its components for joint-stock banks from 2008 to 2017. Source Author’s creation according to annual reports of 12 joint-stock banks

Big-Five showed negative growth. This indicates that in China’s banking industry, the growth of non-interest income is no longer driven solely by the expansion of feebased activities. Indeed, especially with the liberalization of the foreign exchange market, in future there will be more and more space for the development of exchange trading and investment. In addition, from Figs. 2.5 to 2.6 we can observe the similarities and differences in the composition of fee income for different bank types. First, unlike banks in mature markets, Chinese banks do not value the development of consulting and settlement business; consequently, these two businesses do not occupy a high proportion of

38

2 The Rise and Development of Non-traditional Banking Business …

the fee-based business in either of the banking groups. In contrast, bank card and personal wealth management account for relatively high proportions of the fee-based business. Further, while the Big-Five banks maintain a relatively balanced mix of business components, the joint-stock banks prefer to earn additional fees and income through the development of bank card business. From the above it can be seen that there has been a rapid increase in non-interest banking activities, and the proportion of non-interest income in overall operating income is also increasing steadily. Nevertheless, by comparing the figures for the Chinese banking sector with those for mature banking markets, it can be seen that the income diversification process in the Chinese banking sector still lags behind, and there is room for future growth. For example, in 2016 the proportion of non-interest income in the United States banking sector was 32.73%, while the corresponding figure for the European zone was 43.12% (World Bank 2016). In next section, we highlight the reasons why income diversification has become the subject of much positive attention from bank managers, and why the expansion of non-interest income will be the inevitable choice for the Chinese banking market.

2.4 Motivations for Chinese Banks’ Diversification 2.4.1 Constraint-Induced Diversification 2.4.1.1

Market-Oriented Interest Rate Reform

For a long time, interest spread in the Chinese banking sector was rather big which contributed to Chinese banks’ over-reliant on lending business to earn their revenue and interest income in turn constituted the main avenue for their income stream (Ding et al. 2017). As can be seen from Fig. 2.7, the average one-, three-, and fiveyear benchmark interest spreads over the period from 1990 to 1998 were 1.38%, 1.40%, and 1.34%, respectively. The benchmark interest rate reached its highest level in 1999, and has remained at a relatively high level since then. In order to increase the willingness of the banks to diversify their income stream and promote competitiveness in the banking market, in 2005 the government initiated the marketoriented interest rate reform. The reform was introduced gradually. In 2012, the PBOC allowed Chinese financial institutions the freedom to decide on their own deposit interest rates if it was no greater than 10%. The following year, the PBOC opened up the loan interest rate for financial institutions, which further cut the net interest margin for commercial banks. In 2015 the PBOC removed the ceiling it had imposed on deposit rates and abolished the floor for lending rates (Tan et al. 2016). The interest rate marketization process leads to functional changes in Chinese commercial banks (Luo 2017). Interest rate deregulation forces lenders to compete with each other on deposit and loan pricing, which squeezes the net interest margin, or the difference between what lenders pay for deposits and what they collect on

2.4 Motivations for Chinese Banks’ Diversification

39

4.30% 3.30% 2.30% 1.30% 0.30% -0.70% -1.70%

Interest spread one year

Interest spread three years

Interest spread five years

Fig. 2.7 Benchmark interest spreads in China from 1991 to 2015. Notes Interest spread = Benchmark loan rate − Benchmark deposit rate. Source Author’s creation according to the People’s Bank of China

loans (Nguyen 2012; Genay and Podjasek 2014; Alessandri and Nelson 2015). As the main source of profit for Chinese commercial banks, the net interest margin falls to record low (Zuo et al. 2014). According to the Bankscope database, the net interest margin for the Chinese banking sector in 2007 was 3.5%, while by the end of 2017 it had decreased to 2.10%. Such a sizable decrease of net interest margin causes significant shrinkage of banks’ earnings from lending-based activities and of overall profitability ability (Okazaki 2017). The interest rate liberalization has had a large impact on Chinese banks’ profitability (Ding et al. 2017). Figure 2.8 presents the main profitability indexes for 22

50.00%

20

40.00% 30.00%

18

20.00% 16 10.00% 14

0.00%

12

-10.00%

10

-20.00% 2007

2008

2009

2010

2011

Return on equity (right)

2012

2013

2014

2015

2016

2017

Growth rate of profit after tax (left)

Fig. 2.8 ROE and profit after tax in the Chinese banking sector from 2007 to 2017. Source Author’s creation according to China Banking Regulatory Commission

40

2 The Rise and Development of Non-traditional Banking Business …

banks, namely return on equity (ROE) and the growth rate of profit after tax. Both indicators show a similar trend, where following the fluctuation during the subprime mortgage crisis of 2008–2010 in the USA, the profitability of China’s banking industry has been in constant decline. In 2017, the level of ROE reached a low of 12.56, and the growth rate of profitability became negative. Therefore, the Chinese banking sector is faced with a significant challenge posed by declining profitability which also claimed by several researches (such as Bikker and Vervliet 2018). As some non-interest related businesses, such as financial derivatives, securitized products, guarantees and venture capital, can yield high returns, this motivates banks’ management to shift to such business leading to adopting the diversification strategies (Elsas et al. 2010; Dietrich and Wanzenried 2011).

2.4.1.2

Monetary Policy and Capital Requirements

According to Borio et al. (2017) monetary policy has an important effect on banks’ business strategies. Under a loose monetary policy, commercial banks are stimulated to expand their lending businesses, and interest income would dominate their income stream. Conversely, a tight monetary policy would place more restrictions on banks with regard to availability of funds for lending, so income diversification becomes an important alternative for their business opportunity. After the subprime mortgage crisis and in line with the global tendency, the PBOC adopted a moderately loose monetary policy stance and launched a series of measures to stimulate economic growth by injecting liquidity to the financial market (Yang et al. 2017). This however led to excess liquidity in the financial system and rapid growth of banks’ lending. According to CBRC, the total lending volume increased from 27.8 trillion yuan to 42.6 trillion yuan from 2007 to the end of 2009. From 2010, in order to strengthen macro-economic controls, prevent systemic financial risks and avoid increased pace of inflation, the Chinese monetary authorities started to change their monetary policy towards a neutral stance (Xiong 2012). A gradual raising of the deposit reserve ratio and the interest rate restrained the excessive growth of bank loans and led to a decline in the total amount of bank credit. In 2010, in the wake of the European debt crisis, instability in China’s financial market intensified, and the central bank tried to recover excess liquidity in the market by adjusting the excess reserve ratios and open market operations (Jian et al. 2011). In 2012, China’s economic growth slowed down, and the PBOC adopted a “prudent and neutral” monetary policy (Zhang and Sun 2017), thus slowing the expansion of on-balance sheet business and inducing the banks to expand off-balance sheet business. Coupled with this prudent and neutral monetary policy, the Chinese authorities published a series of regulatory rules and notices, aimed at strengthening bank capital requirements and decreasing banks’ overall leverage (Zepeda 2013). In 2012 the CBRC published the Administrative Measures on the Capital of Commercial Banks (Trial), which set a new minimum level of capital indicators, requiring each bank’s core tier one capital ratio, the tier one capital ratio and the capital adequacy ratio to be no lower than 5%, 6% and 8%, respectively. Additionally, commercial banks

2.4 Motivations for Chinese Banks’ Diversification

41

were required to build up a corresponding capital buffer above their minimum capital requirements and reserve capital requirements, including retained capital and countercyclical capital requirements. For the systemically important banks, they were subject to an additional 1% capital requirement. In short, regulatory changes in China have resulted in banks seeking other revenue opportunities beyond those from traditional business. New regulations, particularly the capital requirements, affect banks’ capital availability as well as its cost. New and stricter capital adequacy regulations mean that banks have to raise more capital to meet the capital adequacy requirements (Zhang et al. 2008). This in turn would force banks to seek new capital or to refrain from expanding their traditional banking business. In addition, new regulatory changes would also increase banks’ funding cost, since fewer insider loans would now be available. The risk adjusted capital requirements would also create an incentive for banks to reduce their risk-weighted assets and seek income through non-traditional business (Cohen and Scatigna 2016). Hence, the capital adequacy rules induce banks to seek income from other sources, mainly from non-interest business, which has lower reserves requirements and hence lower capital cost than are associated with traditional lending. Consequently, banks shift to off-balance activities that can generate non-interest income.

2.4.1.3

The Internationalization of the Renminbi

Following the US subprime mortgage crisis, the Chinese authorities accelerated the process of internationalization of the renminbi, in order to bring relief to a monetary system that was too strongly tied to the US dollar (Dobson and Masson 2009). On one hand, the fluctuation of exchange rates of major international currencies, such as the US dollar, euro, and yen, means that Chinese enterprises face significant exposure to exchange rate risks in their international operations. With increased use of the renminbi in international transaction, the RMB internationalization provides Chinese banks with new opportunities for wider engagement in international business and improved income structure in terms of currency exposure (Cohen 2012). Potential new RMB businesses such as overseas banking services in RMB deposits and loans, overseas RMB cash management, currency exchange, international bank cards, and account management will result in large fee income. Moreover, RMB internationalization has the potential of expanding the scale of Chinese banks’ international clearing businesses, which will facilitate trade financing and development of financial product chains (Eichengreen and Kawai 2014). As reported by the annual reports of the Bank of China (2016, 2017), its branches at Hong Kong, Macau and Taiwan have seen a rapid increase in their non-interest income due to the surge in cross-border settlements (see Table 2.2).

42

2 The Rise and Development of Non-traditional Banking Business …

Table 2.2 Accounting indicators of Bank of China, 2015 and 2016 Mainland China

Hong Kong, Macau and Taiwan

Other countries

2015

2015

2015

2016

2016

2016

Assets

13,053.1

14,341.7

3,010.9

3,256.5

1,819.8

1,812.5

Liabilities

11,970.9

13,198.4

2,784.0

2,967.6

1,770.8

1,757.5

Operating income

382.3

365.9

75.2

101.7

17.8

19.1

Net interest income

282.1

263.6

31.7

29.3

14.7

13.0

Non-interest income

100.2

102.3

43.5

72.3

3.0

6.1

Fee and commissions

75.2

70.7

14.7

14.4

3.3

4.2

Other non-interest income

24.9

31.6

28.7

57.9

−0.2

1.8

Source Author’s creation according to yearbooks of Bank of China (2015 and 2016)

2.4.1.4

Reform and Opening up in the Chinese Banking Sector

1. Opening up to foreign capital Since 2001, China joint the WTO, China has committed to opening up its banking sector to foreign capital and investors. In 2006, in accordance with China’s commitment to join the World Trade Organization (WTO), foreign banks was legally granted access to the country’s banking industry (Tsai et al. 2014). Under the 2006 Regulations on Administration of Foreign-Funded Banks, foreign banks operating in China are no longer subject to geographical and business restrictions which causing a rapid increase of foreign banks entered into Chinese banking market (Luo et al. 2015). At the end of 2006, about 30 foreign financial institutions have invested over 19 billion USD into 21 Chinese commercial banks and hold their stakes (Okazaki 2007). According to the latest National Bureau of Statistics, except for the decrease in total assets of foreign-funded banks in 2015, the total assets of foreign-funded banks have generally maintained a significant upward trend, and the average annual compound growth rate of assets for the 10 years reached 10.57% (National Bureau of Statistics China). By the end of 2016, 37 wholly foreign owned banks had established in China; 68 foreign banks had set up 121 branches, and 145 foreign banks had set up 166 representative offices. The total number of business outlets of foreign banks reached 1031 and their total assets reached 2.93 trillion RMB, tripling the amount in 2007 of 927.9 billion RMB. The entry of foreign banks has promoted competitiveness in the Chinese banking sector and exerted a positive effect on domestic banks’ efficiency and their motivations for seeking out new income sources (Claessens et al. 2001; Unite and Sullivan 2003; Choi and Hasan 2005). Moreover, the foreign banks also holds better assets quality and better capital solvency ability (see Fig. 2.9). Thus the entry of foreign banks also brings advanced risks-management technology to domestic banks and causing positively spillover effect (Lee and Hsieh 2014).

43

5.00%

23.00%

4.00%

18.00%

3.00%

13.00%

2.00%

8.00%

1.00%

3.00%

0.00%

-2.00% 2010

2011

2012

2013

2014

2015

2016

Capital adequacy ratio

NPLs ratio

2.4 Motivations for Chinese Banks’ Diversification

2017

Ratio of NPLs Average of Chinese banking sector Ratio of NPLs Foreign bank Capital adequacy ratio Average of Chinese banking sector Capital adequacy ratio Foreign bank

Fig. 2.9 The NPL ratio and capital adequacy ratio of average Chinese banks and foreign banks from 2010 to 2017. Source Author’s creation according to China Banking Regulatory Commission and annual report of each foreign banks

2. Opening up to private capital Compared with the opening up of the banking market to foreign capital, the easing of restrictions on private capital came relatively late but also increased competition for deposits and putting pressure in Chinese banking market (Hou et al. 2016). In 2005, the State Council promulgated the Several Opinions of the State Council on Encouraging, Supporting and Guiding the Development of Individual, Private and Other Non-public Sectors of the Economy. This allowed non-publicly-owned capital to enter the financial services industry for the first time to stimulate its growth and stability (Milana and Wang 2013). Specifically, it permitted private capital to enter regional joint-stock commercial banks and cooperative financial institutions. In 2012, the State Council announced a scheme of Encouraging and Guiding the Entry of Private Capitals in the Fields of the Bank Industry. With that, private capital could enter the banking market, which represents an important progress easing the restrictions on the country’s state-controlled banking industry. In 2014, the first five private banks gained approval for trial operation (see Table 2.3). These private banks offer financing services specifically to small- and medium-sized enterprises (SMEs), self-employed individuals, and others in special development projects such as the Shanghai and Tianjin Pilot Free Trade Zones. At the end of 2016, the total assets of these five private banks had reached 132.9 billion RMB. This bodes well for promoting competition in the banking market (Clarke et al. 2005), but it also casts a shadow on the profitability of the major lenders. Such private bank and internet finance can significantly reduce the transaction costs and information asymmetry. Large and medium lenders will probably experience a gradual outflow of depositors if they do not respond to competitive pressure from smaller banks willing to offer higher deposit rates to win retail clients and from innovative Internet financing platforms offered by the likes of Alibaba and Tencent (Wei 2015).

44

2 The Rise and Development of Non-traditional Banking Business …

Table 2.3 The first batch of five pilot private banks Bank names

Main sponsors

Regions

Customer direction

Permission received

Zhejiang E-Commerce Bank

Alibaba Group and Fosun Group

Zhejiang Province

e-bank

26/92014

Shanghai Huarui Bank

JuneYao Group

Shanghai

Enterprises in Shanghai Pilot Free Trade Zone

26/9/2014

WeBank

Tencent Holdings Ltd.

Shenzhen

e-bank

25/72014

Kincheng Bank of Tianjin

Tianjin Huabei Group

Tianjin

Enterprises in Tianjin Pilot Free Trade Zone

25/72014

Wenzhou Minshang Bank

Chint Group

Wenzhou

small-sized enterprises and rural areas

25/72014

Source Author’s creation according to China Banking Regulatory Commission

In summary, a more flexible entry mechanism is pushing the banking sector to launch more financial products in order to compete with private and foreign banks to attract customers and market share. At the same time, it incentivizes banks to diversify their portfolios in order to look for new profit growth opportunity so that they can move away from the traditional model under which the net interest margin was shrinking.

2.4.2 Internal-Bank Motivations 2.4.2.1

Insufficient Liquidity Due to Resource Misallocation

Over the last two decades, the total assets of the Chinese banking sector have grown rapidly, increasing over nine-fold from 2003 to 2016 (CBRC 2017). However, this rapid growth has brought with it the problem of financial frictions and capital misallocation (Lai et al. 2016), where banks put most of their focus on excessive expanding of the loan scale and lavish local branches, thus causing their left with insufficient liquidity. As can be seen from Fig. 2.10, the overall level of excess reserves in the Chinese banks shows a downward trend with the exception that rural banks are able to maintain an average ratio of excess reserves over 9%. Lately, with the tempering of the US quantitative easing process, the lack of liquidity issue increasingly becomes a problem for Chinese banks and in response, managers of these banks then engage in more diversification activity to increase non-interest income. This to some extent

2.4 Motivations for Chinese Banks’ Diversification

45

14.00% 12.00% 10.00% 8.00% 6.00% 4.00% 2.00% 0.00% 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 overall

Large commercial banks

Joint stock commercial banks

Rural commercial banks

Fig. 2.10 Excess reserves in the Chinese banking sector from 2003 to 2017. (The People’s Bank of China did not report the excess reserves for large and joint stock banks in 2010). Source Author’s creation according to the People’s Bank of China

redresses misallocation of banks’ resources and reduce the capital that is taken up by traditional businesses, thus increasing banks’ liquidity level. The diversification also provides more financing channels for the banks. Consequently, diversification could help increase banks’ short-term capital stocks and make them better able to resist the tightening of liquidity (Pana et al. 2010), thus helping alleviate misallocation of capital and easing banks’ financial distress.

2.4.2.2

Financial Disintermediation

In an increasingly rigorous regulatory environment, banks are faced increased competitions resulting from the squeeze induced by more stringent capital requirements for traditional business (Li et al. 2014). This situation was worsened by the subprime mortgage crisis, which led to banks becoming more prudent in their loan decisions (Jun 2012). Figure 2.11 presents the growth of non-governmental financing in China, and its structure; it shows the percentages of China’s direct and indirect financing in aggregate non-governmental funding. While in 2004 loans accounted for nearly 80% of the total, since then the share has gradually shrunk, reaching its lowest at 51.35% in 2013. During the same period, the growth rate of direct financing reached 108.90% in 2007, and the total volume of direct financing increased nearly seven-fold, from 595.6 billion yuan in 2004 to 4136.9 billion yuan in 2016. The rise of direct financing has significantly eroded the traditional operations that generate interest income and has pushed Chinese banks to diversify their business. Such financial disintermediation is mainly driven by the substantial development of financial markets including that of the stock market, bond market, money market, and gold market, where it creates the opportunity for effective and low-cost financing. The direct financing through financial markets competes with banks’ traditional interest-based activities (Perera et al. 2014). In 2004 the combined market value

46

2 The Rise and Development of Non-traditional Banking Business …

100%

120.00% 100.00%

80% 80.00% 60%

60.00% 40.00%

Equity Financing on the Domestic Stock Market by Non-financial Enterprises Net Financing of Corporate Bonds

40% 20.00% 20%

0.00% -20.00%

Undiscounted Corporate Bankers Acceptance Notes Trust Loans

0% -40.00% -20%

-60.00%

Fig. 2.11 Growth of non-governmental financing and its structure from 2003 to 2015. Source Author’s creation according to National Bureau of Statistics of China

of the Shanghai and Shenzhen stock exchanges was 3705.56 trillion yuan, with the Shenzhen stock market occupying 1104.2 trillion yuan and the Shanghai stock market 2601.43 yuan. At the end of 2017, the combined market value had grown to 56708.6 trillion yuan, an over 15-fold increase (see Fig. 2.12). Therefore, Li and Zhang (2013) suggest that the financial disintermediation in turn squeezes the scale of bank lending, and leads to the increases in the pressure on banks’ traditional business. Consequently, financial disintermediation and the development of the capital market have promoted the transfer, sale, and securitization of loans and have driven banks to diversify their income stream, change their business modes, and expand their non-interest income. In addition, financial disintermediation promotes product innovation, capital settlement, asset custody, investment banking, and capital 60,000.00

Trillion CNY

50,000.00 40,000.00 30,000.00 20,000.00 10,000.00 0.00

Shenzhen Stock Exchange

Shanghai Stock Exchange

Fig. 2.12 Capitalisation of the Chinese stock markets. Source Author’s creation according to Shanghai stock exchange and Shenzhen stock exchange

2.4 Motivations for Chinese Banks’ Diversification

47

market-related businesses, all of which are beneficial to banks that wish to provide diversified investment and financing services for corporate clients.

2.4.2.3

Non-performing Assets

Billion CNY

In response to the fallout of the global financial crisis, China’s top economic planner, the National Development and Reform Committee (NDRC), launched a stimulating programme worth of 586 billion USD, with most of the funds being invested in large-scale infrastructure projects and industrial restructuring (Lee 2009). Although this eased the impact of the global financial crisis on China, it also raised concerns for the possibility of reckless lending by banks, since the process would be full of government interventions, which would not only distort the efficient allocation of capital but wold also fuel the zealous of the local governments for channelling the funds to the projects in their localities (Chen et al. 2017). This would create the situation in which local governments’ finance is over-stretched. In the real economy, it also intensifies manufacturing over-capacity in the industries from shipbuilding to solar energy, threatening occurrence of large numbers of non-performing bank loans (Wang 2011). During the recent financial crisis and in the years leading up to it, the Chinese government injected significant amounts of capital into the Chinese banking market in order to write off substantial bad loans and thus create a more healthy level of non-performing loans (NPLs) (Dobson and Kashyap 2006; Tan and Floros 2013; Fu et al. 2015). According to the CBRC, the value of NPLs remained stable during the period from 2008 to 2013, fluctuating only slightly within a range from 400 to 500 billion CNY (see Fig. 2.13). From 2014, there were signs of a rebound in NPLs, due to the slowdown in the macro-economy. By the end of 2014, the total value of NPLs had reached 842.56 billion CNY, an increase of 348.71 billion CNY on the 1600 1400 1200 1000 800 600 400 200 0

1.80% 1.60% 1.40% 1.20% 1.00% 0.80% 0.60% 0.40% 0.20% 0.00% 2010

2011

2012

2013

2014

2015

Non-performing loan volume

Substandard volume

Doubtful volume

Loss volume

2016

Non-performing loan rate

Fig. 2.13 Non-performing loans by Chinese banks from 2010 to 2016. Source Author’s creation according to China Banking Regulatory Commissions annual reports

48

2 The Rise and Development of Non-traditional Banking Business …

previous year. By the end of 2016, the outstanding bad loans had increased to 1.51 trillion yuan, up 19% from a year earlier. As shown in Fig. 2.13, the growth rate of non-performing loans increased from −12.81 to 50.72% over the period from 2010 to 2016. The situation prompted the banks to search for new income possibilities, including non-traditional activities.

2.5 Conclusion The chapter provides an overview of the rise of non-interest activity in the Chinese banking industry against the background of China’s evolving financial reform and the regulatory changes thereof. Through a process of reform carried out in three waves, the Chinese authorities established a modern banking system. The current Chinese financial system features a framework with “one bank and two commissions”. In which the central bank plays a central role in setting the monetary policy while the main regulatory functions are charged to two regulatory commissions (Zhu and Hu 2019). Compared to the mature banking market, income diversification by Chinese banks is still at an early stage, and net interest income remains the dominant source of their revenues, occupying 73.4% of the overall Chinese banking income stream (China Banking Regulatory Commission). The proportion of non-interest income in the total revenue of Chinese banks remains relatively low. However, the volume of non-interest income is increasing rapidly (Sun et al. 2017). Changes in the income structure of China’s banking industry are driven mainly by development of financial reforms, deepening financial disintermediation and the process of interest rate marketization (Li and Zhang 2013). In addition, exchange rate reform and RMB internationalization have provided Chinese banks with new opportunities for wider engagement in international business and improved income structure in terms of currency exposure (Cohen 2012). China’s opening up to private and foreign capital, and the increasingly stringent regulation on capital requirements, have increased competition for deposits and put pressure on the wider Chinese banking market. Consequently, they have played a significant role in driving Chinese banks to shift from traditional banking to non-interest activity (Borst and Lardy 2015; Hou et al. 2016). China’s banks is faced with enormous challenges ahead. The unfolding of financial reforms, tighter bank regulatory rules and economic uncertainties in the world and domestic economy will continue to squeeze bank profits and increase their exposure to risk. In the event, it is imperative that Chinese banks would have to adopt a sound diversification strategy and pursuit mixed business lines to deal with the challenges.

References

49

References Alessandri P, Nelson BD (2015) Simple banking: profitability and the yield curve. J Money Credit Bank 47(1):143–175 Almanac of China’s Banking and Finance (1993) People’s Publishing House:30–59 Bank of China Annual Report (2016):55–70 Bank of China Annual Report (2017):30–68 Berger AN, Hasan I, Zhou M (2009) Bank ownership and efficiency in China: what will happen in the world’s largest nation? J Bank Financ 33(1):113–130 Bikker JA, Vervliet TM (2018) Bank profitability and risk-taking under low interest rates. Int J Financ Econ 23(1):3–18 Bin LPW (2007) The shareholding system reform of state-owned commercial banks: a view of institutional economics. Nankai Econ Stud 3:20–35 Bonin JP (1999) Banking reform in China: gradually strengthening pillar or fragile reed?. Wesleyan University. Working paper Borio C, Gambacorta L, Hofmann B (2017) The influence of monetary policy on bank profitability. Int Financ 20(1):48–63 Borst N, Lardy N (2015) Maintaining financial stability in the People’s Republic of China during financial liberalization. Peterson Institute for International Economics. Working paper Boubakri N, Cosset JC, Guedhami O (2005) Liberalization, corporate governance and the performance of privatized firms in developing countries. J Corp Financ 11(5):767–790 Boyreau-Debray G, Wei SJ (2005) Pitfalls of a state-dominated financial system: the case of China. Natl Bur Econ Res 35–47 Cai Y, Hao X, Han E, Gao K (2019) Financialization, monetary policy and technological innovation. Sciences 7(1):1–13 Chen CR, Huang YS, Zhang T (2017) Non-interest income, trading, and bank risk. J Financ Serv Res 51(1):19–53 China Banking Regulatory Commission Annual Report (2016):28–60 China Banking Regulatory Commission Annual Report (2017):43–55 Choi S, Hasan I (2005) Ownership, governance, and bank performance: Korean experience. Financ Mark Inst Instrum 14(4):215–242 Claessens S, Demirgüç-Kunt A, Huizinga H (2001) How does foreign entry affect domestic banking markets? J Bank Financ 25(5):891–911 Clarke GR, Cull R, Shirley MM (2005) Bank privatization in developing countries: a summary of lessons and findings. J Bank Financ 29(8–9):1905–1930 Cohen BH, Scatigna M (2016) Banks and capital requirements: channels of adjustment. J Bank Financ 69:S56–S69 Cohen BJ (2012) The yuan tomorrow? evaluating China’s currency internationalisation strategy. New Polit Econ 17(3):361–371 Deng Y, Luo Q (2014) A Study of the strategy of RMB internationalization in the context of financial crisis. Contemp Logist 15:103–115 Dietrich A, Wanzenried G (2011) Determinants of bank profitability before and during the crisis: evidence from Switzerland. J Int Financ Mark Inst Money 21(3):307–327 Ding N, Fung HG, Jia J (2017) Comparison of bank profitability in China and the USA. China and World Econ 25(1):90–108 Dobson W, Kashyap AK (2006) The contradiction in China’s gradualist banking reforms. Brook Pap Econ Act 2:103–162 Dobson W, Masson PR (2009) Will the renminbi become a world currency? China Econ Rev 20(1):124–135 Eichengreen B, Kawai M (2014) Issues for renminbi internationalization: an overview. ADBI Working Paper No.454 Elsas R, Hackethal A, Holzhäuser M (2010) The anatomy of bank diversification. J Bank Financ 34(6):1274–1287

50

2 The Rise and Development of Non-traditional Banking Business …

Fang X, Jiang Y (2014) The promoting effect of financial development on economic growth: evidence from China. Emerg Mark Financ Trade 50(sup1):34–50 Fu Y, Lee SC, Xu L, Zurbruegg R (2015) The effectiveness of capital regulation on bank behavior in China. Int Rev Financ 15(3):321–345 García-Herrero A, Gavilá S, Santabárbara D (2009) What explains the low profitability of Chinese banks? J Bank Financ 33(11):2080–2092 Genay H, Podjasek R (2014) What is the impact of a low interest rate environment on bank profitability?. Chicago Fed Letter Hamid J, Tenev S (2008) Transforming China’s Banks: the IFC’s experience. J Contemp China 17(56):449–468 Hasan I, Wachtel P, Zhou M (2009) Institutional development, financial deepening and economic growth: evidence from China. J Bank Financ 33(1):157–170 He L, Chen L, Liu FH (2017) Banking reforms, performance and risk in China. Appl Econ 49(40):3995–4012 Hou X, Gao Z, Wang Q (2016) Internet finance development and banking market discipline: evidence from China. J Financ Stab 22:88–100 Hsiao C, Yan S, Wenlong B (2015) Evaluating the effectiveness of China’s financial reform—The efficiency of China’s domestic banks. China Econ Rev 35:70–82 Jia C (2009) The effect of ownership on the prudential behavior of banks: the case of China. J Bank Financ 33(1):77–87 Jian H, Shaoyi C, Yanzhi S (2011) Capital inflows and the impossible trinity in China. J Int Glob Econ Stud 4(2):30–46 Jiang C, Yao S, Feng G (2013) Bank ownership, privatization, and performance: evidence from a transition country. J Bank Financ 37(9):3364–3372 Jun TL (2012) China financial disintermediation in late finance crisis time and commercial bank’s countermeasures. J Spec Zone Econ 4:021 (in Chinese) Laeven L, Levine R (2007) Is there a diversification discount in financial conglomerates? J Financ Econ 85(2):331–367 Lai TK, Qian Z, Wang L (2016) WTO accession, foreign bank entry, and the productivity of Chinese manufacturing firms. J Comp Econ 44(2):326–342 Lee CC, Hsieh MF (2014) Bank reforms, foreign ownership, and financial stability. J Int Money Financ 40:204–224 Lee CH (2009) National policy responses to the financial and economic crisis: the case of China. ILO Regional Office for Asia and the Pacific, No, p 22 Li J, Hsu S, Qin Y (2014) Shadow banking in China: institutional risks. China Econ Rev 31:119–129 Li L, Zhang Y (2013) Are there diversification benefits of increasing noninterest income in the Chinese banking industry? J Empir Financ 24:151–165 Liao W, Tapsoba MSJA (2014) China’s monetary policy and interest rate liberalization: lessons from international experiences. Int Monet Fund 14–75 Lin JY, Sun X, Wu HX (2015) Banking structure and industrial growth: evidence from China. J Bank Financ 58:131–143 Lin X, Zhang Y (2009) Bank ownership reform and bank performance in China. J Bank Financ 33(1):20–29 Lo K, Xue L, Wang M (2016) Spatial restructuring through poverty alleviation resettlement in rural China. J Rural Stud 47:496–505 Lozano-Vivas A, Pasiouras F (2014) Bank productivity change and off-balance-sheet activities across different levels of economic development. J Financ Serv Res 46(3):271–294 Luo J (2017) Shadow banking, interest rate marketization and bank risk-taking: an empirical study of the 40 commercial banks in China. J Financ Risk Manag 6(1):27–36 Luo D, Yao S (2010) World financial crisis and the rise of Chinese commercial banks: an efficiency analysis using DEA. Appl Financ Econ 20(19):1515–1530 Luo Y, Qian X, Ren J (2015) Initial public offerings and air pollution: evidence from China. J Asia Bus Stud 9(1):99–114

References

51

Luo D, Dong Y, Armitage S, Hou W (2017) The impact of foreign bank penetration on the domestic banking sector: new evidence from China. Europ J Financ 23(9):752–780 Milana C, Wang J (2013) Fostering entrepreneurship in China: a survey of the economic literature. Strat Chang 22(7–8):387–415 Nguyen J (2012) The relationship between net interest margin and noninterest income using a system estimation approach. J Bank Financ 36(9):2429–2437 Okazaki K (2007) Banking system reform in China: the challenges of moving toward a marketoriented economy. RAND’s Natl Secur Res Div 194 Okazaki K (2017) Banking system reform in China: the challenges to improving its efficiency in serving the real economy. Asian Econ Policy Rev 12(2):303–320 Pana E, Park J, Query T (2010) The impact of bank mergers on liquidity creation. J Risk Manag Financ Inst 4(1):74–96 Peng J, Groenewold N, Fan X, Li G (2014) Financial system reform and economic growth in a transition economy: the case of China, 1978–2004. Emerg Mark Financ Trade 50(sup2):5–22 Perera A, Ralston D, Wickramanayake J (2014) Impact of off-balance sheet banking on the bank lending channel of monetary transmission: evidence from South Asia. J Int Financ Mark Inst Money 29:195–216 Siregar HS, Sadalia I, Chandra G (2017) The effect of Lerner index and income diversification on the general bank stability in Indonesia. Banks Bank Syst 12(4):56–64 Stiroh KJ, Rumble A (2006) The dark side of diversification: the case of US financial holding companies. J Bank Financ 30(8):2131–2161 Sun L, Wu S, Zhu Z, Stephenson A (2017) Noninterest income and performance of commercial banking in China. Sci Program 10–25 Tan Y, Floros C (2013) Risk, capital and efficiency in Chinese banking. J Int Financ Mark Inst Money 26:378–393 Tan Y, Ji Y, Huang Y (2016) Completing China’s interest rate liberalization. China World Econ 24(2):1–22 Tsai YJ, Chen YP, Lin CL, Hung JH (2014) The effect of banking system reform on investment–cash flow sensitivity: evidence from China. J Bank Financ 46:166–176 Unite AA, Sullivan MJ (2003) The effect of foreign entry and ownership structure on the Philippine domestic banking market. J Bank Financ 27(12):2323–2345 Wang H (2018) China and international financial standards: from “rule taker” to” rule maker”?. CIGI Working Paper Wang K, Huang W, Wu J, Liu YN (2014) Efficiency measures of the Chinese commercial banking system using an additive two-stage DEA. Omega 44:5–20 Wang Y (2011) China and the United States’ recovery from the global financial crisis. Unpublished manuscript Wei S (2015) Internet lending in China: status quo, potential risks and regulatory options. Comput Law Secur Rev 31(6):793–809 Xiong W (2012) Measuring the monetary policy stance of the People’s bank of China: an ordered probit analysis. China Econ Rev 23(3):512–533 Yang Z, Wu S, Shen Y (2017) Monetary policy, house prices, and consumption in China: a national and regional study. Int Real Estate Rev 20(1):23–49 (in Chinese) Yao S, Han Z, Feng G (2008) Ownership reform, foreign competition and efficiency of Chinese commercial banks: a non-parametric approach. World Econ 31(10):1310–1326 Yao Y, Wu HM (2010) History, politics and 30 years of development and reform development and reform. In: Reform and development in China, pp 65–90 Zepeda R (2013) Pioneering a new legal and regulatory framework for derivatives markets in China: the third way. J Int Bank Law Regulat 28(7):251–265 Zhang C, Sun Y (2017) Confidence in Chinese monetary policy. Int Rev Econ Financ 52:212–221 Zhang D, Cai J, Dickinson DG, Kutan AM (2016) Non-performing loans, moral hazard and regulation of the Chinese commercial banking system. J Bank Financ 63:48–60

52

2 The Rise and Development of Non-traditional Banking Business …

Zhang J, Jiang C, Qu B, Wang P (2013) Market concentration, risk-taking, and bank performance: evidence from emerging economies. Int Rev Financ Anal 30:149–157 Zhang ZY, Jun WU, Liu QF (2008) Impacts of capital adequacy regulation on risk-taking behaviors of banking. Syst Eng-Theory Pract 28(8):183–189 (in Chinese) Zhu CC, Hu J (2019) Internet financial regulation and law analysis of China. In: Green Finance for Sustainable Global Growth, pp 173–192 Zuo Z, Tang XG, Liu YZ (2014) Will the deposit interest rate marketization increase bank’ risk level?: based on an angle of narrowed gap between deposit and loan interest rate. Financ Econ 2:20–29 (in Chinese)

Chapter 3

The Impact of Income Diversification on Chinese Banks: Bank Performance

Abstract This chapter investigates the effects of income diversification on the performance of banks in China. By adopting the system-GMM estimation, this research finds the different diversification impacts on performance for three categories of banks. By decomposing non-interest activities into different components, it finds further significant results.

Chapter 2 has described and analysed the major structural reforms undergone by the Chinese banking industry, as well as the dramatic developments in terms of macroeconomics, financial liberalization, capital restriction and internal operational pressures on banks. These have caused huge changes in both the banks’ income structure and its components, where non-interest income is increasing in amount and in proportion to banks’ assets and operating income. Against this background of structural change, in this chapter will investigate whether income diversification in Chinese banks results in better earnings and overall performance.

3.1 Research Background Regulatory changes and banking competition in recent decades have brought about significant changes to the banking market, including the function of financial institutions and their income structure (Allen and Santomero 2001). In this changing environment, banks are actively seeking new income streams and business lines, as opposed to the more traditional interest margin income (Casu et al. 2016). Hence, there has been a continual increase in banks’ non-interest activities, leading banks to diversify from traditional interest-bearing loans to earning income from offering a broad range of mixed financial products and services. However, evidence regarding the impact of income diversification on bank performance has been inconclusive. Some claim that diversification is beneficial to banks since business expansion to non-interest activities can increase banks’ market power and competitive advantages and can offer lower financing cost. On the other hand, © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0_3

53

54

3 The Impact of Income Diversification on Chinese Banks …

15.00%

20

10.00%

10

5.00%

0

0.00%

Trillion CNY

30

25 20 15 10 5 0

15.00% 10.00% 5.00% 0.00%

Non-Interest Operating Income

Fee and Commissions

Share of Non-Interest Operating Income

Share of Fee and Commissions

2 1.5 1 0.5 0 -0.5 -1

2.50% 2.00% 1.50% 1.00% 0.50%

Trillion CNY

Trillion CNY

Trillion CNY

there are opposite views that banks with more diversified portfolios are also likely to perform lower well than traditional institutions. Income diversification can also create the presence of too many business lines and disorganized management, offsetting benefits such as supernormal returns and competitiveness. Therefore, whether income diversification can bring a benefit or discount calls for more evidence. Existing studies have been mostly concerned with mature markets. Only a small though expanding body of the literature has focused on emerging economies. Given their growing importance in international finance, it is desirable to consider how income diversification fares in a wide range of countries, including China. As the China is the largest emerging economy with global importance, studies on China would shed further light on the issue of bank diversification and its consequences. In the face of the increasingly stringent regulatory capital standards, Chinese banks are under pressure to seek new sources of funding. Meanwhile, market-oriented interest rate reform and financial disintermediation have gradually narrowed banks’ net interest margins, encouraging banks to diversify their income stream. Figure 3.1 illustrates the tendency concerning Chinese banks’ shares of interest and non-interest incomes in total operating income for 2005–2016. Figure 3.1 shows the amount and share of non-interest income over total operating income, as well as the three underlying activities, including fee-based, trading and other non-interest income. Overall, there is a continually increasing tendency of total non-interest income in the Chinese banking industry; Fig. 3.1 also indicates

2

1.50%

1.5

1.00%

1 0.50%

0.5 0

0.00%

0.00% Trading and Derivatives

Other Operating Income

Share of Trading and Derivative

Share of Other Operating Income

Fig. 3.1 Chinese banks’ non-interest income. Notes Operating income = Net interest income + non-interest income. Source Author’s calculations based on BankScope database

3.1 Research Background

55

that the main income source of non-interest income is fee activities, which accounts for 12.04% of total non-interest income, while the share in 2016 was only 2.59%. Meanwhile, the net operating revenue of trading and other operating income has experienced some fluctuations. Both suffered a decreasing tendency from 2005 to 2007, and then, they dramatically increased in 2008, with the shares then becoming stable from 2008 onward (growing annually by about 0.5% and 0.8%, respectively). This chapter employs a sample of 40 Chinese commercial banks, which accounts for 79% of the total assets of the Chinese banking industry. Following the classification of the China Banking Regulatory Commission (CBRC), this sample is classified into three groups: global systemically important banks (G-SIBs), domestic systemically important banks (D-SIBs), and other banks that are not classified by the authorities as systemically important (N-SIBs). Dynamic panel data models are employed in this chapter to assess banking groups’ performance in relation to diversification. The discovered evidence suggests that in the Chinese banking industry, income diversification is generally nonprofitable, but the performance effects vary among G-SIBs, D-SIBs and N-SIBs. G-SIBs exhibit the strongest income diversification benefits, while the performance response of D-SIBs is non-significant. The N-SIBs, however, have a significant diversification discount. Generally speaking, there are three main factors that undermine the robustness and clarity of previous research in this field. First, many studies refer to a time period before 2005 (e.g. Berger et al. 2010; Li and Zhang 2013). However, it was not until 2005 that the Chinese regulatory authorities launched the pilot program allowing cooperation among financial institutions, which increased the complexity and diversification of banks’ income streams. In practice, therefore, the process of income diversification by Chinese banks began in 2005, while the non-interest income before that date was mainly from the fee-based business derived from interest income. Moreover, as Lou (2008) points out, before 2005 non-interest income occupied only a very small proportion of operating income, hence the relationship between noninterest income and banks’ performance or risk indicators might not have been in evidence, but would be fully generated only from the expansion of interest income. For these reasons, the results of studies that take into account the period before 2005 must be questionable in terms of their accuracy and robustness. Secondly, most of the existing research in this field uses the pooled OLS estimator, especially in the Chinese case (e.g. Stiroh and Rumble 2006; Baele et al. 2007; Gamra and Plihon 2011). However, those studies have noted a potential endogeneity problem, that is, a possible two-way correlation where risky banks might be more likely to expand their diversification to an extreme level and several factors such as business opportunities and competition levels might affect both dependent and independent variables (e.g. Acharya et al. 2006; Stiroh and Rumble 2006; Baele et al. 2007). Therefore, the use of OLS estimation might allow the introduction of bias, so that it is necessary to control for the endogeneity problems in relation to the diversification process and its conseques. Therefore, this study adopts a dynamic model and compares the results to check whether outcomes differ among several performance indicators. To construct a dynamic model system, GMM (SYS-GMM)

56

3 The Impact of Income Diversification on Chinese Banks …

for example is used as the econometric model, as it can also solve the endogeneity problem associated with OLS. Finally, unlike the interest activities, banks’ non-interest activities are not subject to capital restrictions. Consequently, in an environment with tight capital restrictions, banks have an incentive to develop non-interest businesses that do not use banks’ capital. In other words, the diversification strategy is highly dependent on the capital restrictions. However, this factor is largely neglected by the current literature that seeks to estimate the relationship between Chinese banks’ performance and income diversification, which might mean that the results in those studies are subject to bias. To solve this problem, we categorize the banking industry into three groups, i.e. G-SIBs, D-SIBs and N-SIBs, based on the banks’ systemic importance. In more detail, the Chinese banking regulatory authorities impose different capital restrictions on different types of banks, according to the categories of systemic importance. In 2012, China’s Banking Regulatory Commission published the Administrative Measures on the Capital of Commercial Banks (Trial), which regulates the minimum levels of the core tier-one capital ratio, the tier-one capital ratio and the capital adequacy ratio for G-SIBs, D-SIBs and N-SIBs. According to this provision, the minimum level of core tier-one capital is set at 5%, and the capital adequacy ratio is set at 8%. D-SIBs are required to have an additional 1% risk-weighted assets, which should be satisfied by the core tier-one capital ratio. However, for G-SIBs, the requirement follows the Basel Committee, which stipulates a minimum 8% core tier-one capital ratio and an 11.5% capital adequacy ratio. Investigation of the diversification effects that takes account of this categorization can offer a new perspective and full consideration of the financial environment and financial restriction in China. The remainder of the chapter is organized as follows. Section 3.2 reviews the related literature. Section 3.3 presents details of the data sample, variables and methodology. Section 3.4 reports the empirical estimation and results. Finally, Sect. 3.5 concludes.

3.2 Related Literature Diversification strategy, originally proposed by Ansoff (1957), is a fluid concept with no fixed definition in the literature. While for some researchers, diversification indicates the horizontal boundaries in terms of products, services and markets (Elsas et al. 2010), others refer to the methods used to achieve the goals of business growth and risk reduction (Hoskisson and Hitt 1990). In this paper, we define diversification as the process of conglomeration by mixing business lines within an institution. Diversification can be achieved through geographic diversification, international diversification and income diversification (Mulwa et al. 2015). This chapter is mainly concerned with income diversification, which involves the behaviour whereby banks seek new sources and types of revenue other than traditional interest-bearing loans. Traditional theory arguing for the benefits of diversification is based on the potential that banks may gain benefits through the portfolio effect and economies of

3.2 Related Literature

57

scope (Casu et al. 2016). Given that non-interest incomes are not perfectly related to revenues from traditional financial services, diversification can reduce variations in banks’ returns and profits. Banks also can benefit from economies of scope since non-interest activities are largely based on the branch infrastructure and electronic banking system; they share the initial cost of their traditional business, leading to increased economic scope and benefiting banks (Jagtiani et al. 1995). Diversification as a strategy may help banks gain access to market power and release market competition (Barney 2002). Such competition may in turn bring about efficiency and innovation in the banking industry (Morgan and Samolyk 2003; Landskroner et al. 2005; Lepetit et al. 2008). For banks, a diversified business could help them gain competitive advantages in other financial markets and access to market power owing to cheap capital funding. When banks expand their businesses to noninterest activities, they could increase their market power and competitive advantages since the process would offer better investment opportunities and lower financing cost owing to the portfolio effect and economies of scales. Diversified banks can concentrate funds and invest in less competitive markets to control market prices and prevent potential competitors from entering the industry (Palich et al. 2000). Diversified banks can also obtain rich information from their mixed business lines. The use of the information can help banks overcome information asymmetry when providing traditional lending and improve risk management (Diamond 1984; Ramakrishnan and Thakor 1984; Petersen and Rajan 1995; Stein 2002; Elsas 2005; Drake et al. 2009; Elsas et al. 2010). As banks gain superior informational resources (Massa and Rehman 2008) and superior technological resources (Miller 2004) from diversification, the incorporation of such resources would increase their competitive advantage and capability in other financial activities. However, income diversification might also lead to the presence of too many business lines and to disorganized management, thus creating inefficiency in the internal capital market (De Haas and Van Lelyveld 2010). Moreover, in a deregulated banking system, especially in the case of emerging market countries, resource allocation could be affected by the agency problem, such that higher profitability projects could not claim advantageous resources. In the long term, such inefficiency and misallocation could harm the interest of banks. Empirically, early research tends to be based on simulation exercises (e.g., Boyd et al. 1993, Kwan and Laderman 1999, Lown et al. 2000; Allen and Jagtiani 2000) or on the stock market to determine the effects of diversification (e.g., Lang and Stulz 1994; Comment and Jarrell 1995; Baele et al. 2007). Recently, the accounting approach has become the main method that researchers use to identify the effects of diversification on performance. Relying on balance-sheet data, this approach classifies diversification levels by the proportion of non-interest income. Empirical evidence has been mixed. Landskroner et al. (2005) show that gains from diversification exist. As the scope of banking activities increases, there is a strong positive relation between risk-adjusted performance and asset allocation. Chiorazzo et al. (2008) provide positive evidence suggesting that diversification can improve the trade-off between risk and income for Italian banks. Köhler (2014, 2015)

58

3 The Impact of Income Diversification on Chinese Banks …

finds that banks’ earnings are more stable and profitable if they diversify into noninterest income. Similar results are also reported by Mergaerts and Vander (2016) and Nguyen et al. (2016). However, further literature indicates it is uncertain whether banks can gain a performance improvement from diversification. Jagtiani et al. (1995) provide empirical evidence from the US banking industry showing that non-interest activities have little or no impact on bank costs. Boyd and Runkle (1993) show that larger and diversified banks employ more financial leverage and earn lower profits. DeYoung and Roland (2001) construct a ‘degree of total leverage’ framework and find that diversification has a negative impact on performance, which is echoed in Esho et al. (2005). Stiroh (2004) utilizes a sample of consolidated financial holding companies for the period from 1997 to 2004 and finds that an increase in non-interest income does not lead to higher equity returns. Most of the existing research in the field concerns banks in mature markets. A small but expanding literature has recently emerged to focus on emerging market economies. This body of research generally reports different outcomes than those from mature markets. Khanna and Yafeh (2005) suggest that diversification in emerging market banks leads to only a small discount or premium. Claessens et al. (2001) compare the listed financial groups from the US, Japan and eight East Asian countries, and they find that the level of diversification in East Asian countries is higher and that they maintain a relatively low diversification disadvantage. In their study of seven emerging markets, Lins and Servaes (2002) find a low value discount and further propose that ownership concentration is significantly positively related to the value discount. Sanya and Wolfe (2011) suggest that revenue diversification can also be beneficial to banks in developing countries. The literature focusing on diversification in Chinese banks is limited because diversification is a fairly recent development in China. As in other markets, results from this limited body of literature are mixed. Studies such as Deng and Li (2006), Chi (2006), Zhou and Wang (2008), and Lou (2008) all find some evidence that diversification may improve banks’ performance, but the benefits are often mitigated or even offset by the late development of the move toward diversification and the general lack of skills among bank staff. In the early sample years of these studies (1999–2006), Chinese banks were actually highly specialized. It was only in 2005 that the Chinese regulatory authorities started to pilot a program allowing Chinese banks to engage in non-traditional businesses. Thus, in their sample period, non-interest income accounted for only a small proportion of their total income. Meanwhile, at the time, Chinese banks faced a situation in which the proportion of non-interest income was too low to offer significant benefits, while at the same time, less professional management and lacking experience meant that the cost of non-interest activities was high, which reduced net profits (Lou 2008).

3.3 Variables, Data and Methodology

59

3.3 Variables, Data and Methodology 3.3.1 Variables All the variables in this research are used at a yearly frequency. The definitions of each variable are as follows: (1) Performance Measurements Following Berger and Bouwman (2013), we evaluate banks’ profitability by using the pre-tax returns on both total assets and equity:  ROAit = NIATit  ROEit = NIATit

(Assetit + Assetit−1 ) 2 Equityit + Equityit−1 2

 (3.1)  (3.2)

where ROAit , ROEit and NIATit refer to return on assets, return on equity and net income after tax for bank i in the period t respectively. (2) Risk-Adjusted Performance Measurements Following Stiroh (2004) and Sanya and Wolfe (2011), we construct risk-adjusted returns on both assets and equities. They are the ratios of ROA and ROE for a given year to the standard deviation of ROA and ROE over the sample period: RAROAit = ROAit /σ ROAi

(3.3)

RAROEit = ROEit /σ ROEi

(3.4)

where ROEit and ROAit refer to the pre-tax return on equity and total assets for bank i in the period t respectively. σ ROEit and σ ROAit refer to the standard deviation of the pre-tax return on equity and on total assets respectively. (3) Income Diversification The BankScope database divides operating income into interest and non-interest income. Interest income is sourced from the interest on advantages and investment activities, while all other income is classified as non-interest income. Following Stiroh and Rumble (2006), this paper further divides non-interest income into three components: income from trading in foreign exchange and fiduciary activity; fee and commission income from clearing, settlement and other financial services; and other non-interest income.

60

3 The Impact of Income Diversification on Chinese Banks …

The most commonly employed measure of diversification in the literature is the Herfindahl-Hirschman Index (HHI). In this measure, income diversification is defined as the sum of the square of proportion of individual income sources over total operating income within a bank as follows: HHIi,t = 1 −

n  

xi,t

2

(3.5)

i=1

where n is the number of income categories groups and xi,t measures the category i in period t.. As income can be sourced from interest and non-interest activities, the HHI1 can be described as:  HHI = 1 − (INT/TOR)2 + (FEE/TOR)2 + (TRA/TOR)2 + (OTH/TOR)2 (3.6) where INT is gross interest revenue. According to Elsas et al. (2010), the use of INT can avoid distortions caused by the profitability of a bank’s interest-based business. However, Bankscope and banks’ annual reports do not supply sufficient data on the total income from trading, fees and other activities. As the direct expense for such activities ranges from 5 to 15%, we follow Elsas in calculating the net income from such activities. Thus, in the above HHI measure, TOR describes the total operating revenue; COM refers to the ratio of net fee and commission income to total operating income; TRA is the ratio of net trading income to total operating income, and OTH indicates the ratio of net other operating income to total operating income. TOR describes the total operating revenue, which is the sum of the absolute values of INT, FEE, TRAD and OTH. (4) Other Variables NIM: Net interest margin indicates the net interest revenue over total earning assets. This is intended to describe interest-based activity (Lepetit et al. 2008; Busch and Kick 2009; Köhler 2014). LTA: To assess the correlation between diversification and banks’ lending business, we use the loans-to-assets ratio (LTA) to measure the level of loan investment at the individual bank level (Stiroh 2004; Cornett et al. 2010; Calmès and Théoret 2014). NON: This indicates the ratio of non-interest expenses to total assets. On the one hand, it reflects the efficiency of banks’ cost management, where poor cost management is likely to cause lower bank performance. On the other hand, Busch and Kick (2009) maintain that higher investment can improve the monitoring of borrowers and result in better personnel training ability. Thus, it can reduce the potential loss and improve the capacity to expand non-interest activities.

1 For the sake of consistency, in this book the Herfindahl-Hirschman Index is expressed in percentage.

3.3 Variables, Data and Methodology

61

3.3.2 Data Sample In investigating the impact of income diversification on bank performance, we employ a dynamic panel data model. Yearly panel data are employed, with 40 Chinese banks. All individual bank-level variables are taken from BankScope and individual banks’ own annual reports. The sample period runs from 2005 to 2016. Following the BIS definition and Chinese regulator classification, we divide the sample into three groups, namely, G-SIBs, D-SIBs and N-SIBs, to reflect possible effects of size, managerial efficiency and capital restriction. The criteria for classification as systemically important banks reflect banks’ specific characteristics, which also have a large impact on the effects of banks’ diversification. First, banks classed as G-, D-, and N-SIBs are of significantly different sizes. In 2017, G-SIBs possessed average total assets of 3,406 billion USD, while the average assets of D-SIBs and N-SIBs were 808 and 128 billion USD, respectively. Therefore, there are huge gaps between the three groups in terms of business scale. In larger banks the more extensive non-interest business shares the initial cost of traditional business, and non-interest activities can continue to grow as long as they generate fee income. Moreover, as suggested by Gurbuz et al. (2013), large-sized banks generally have better information technology, human capital management and risk management. Therefore, such business expansion could improve the overall productivity and cause technology spillover within the banking system (Canals 1994; Acharya et al. 2002; Mercieca et al. 2007). Secondly, in the Chinese banking sector the three groups follow significantly different diversification strategies. For G-SIBs, the diversification process is largely influenced by government intervention. Banks in this category are responsible for piloting China’s financial reform, and lead the industry in terms of the scale and expansion of non-interest activities. In contrast, because D-SIBs operate nationally, and are chiefly concerned with competing with G-SIBs to attract customers and deposits to their interest business lines, banks in this group tend to be less diversified. Meanwhile, the much smaller N-SIBs face capital restrictions and have weak risktaking capability. Consequently, banks in this group engage in much lower levels of non-interest business than do G-SIBs and D-SIBs. Thirdly, the three banking categories face different levels of capital regulations. In general, banks in the N-SIBs group must maintain a minimum 8% of the capital adequacy ratio, while for D-SIBs the minimum is 11.5% and G-SIBs are required to hold an additional 1% risk-weighted assets. As suggested by Danila (2013), regulatory restrictions could impact on banks’ traditional activities, as they change the banks’ risk preference and willingness to fund loans. Specifically, in order to keep deposit resources fully invested and allocated, banks tend to allow a larger proportion of higher-risk credit. Such a policy changes banks’ portfolio structure and reduces the quality of assets, making banks and financial systems subject to greater volatility, and causing instability. If managers can originate loans of one default risk efficiently, they will be minimum-cost originators of loans across a spectrum of default risks. This ability to originate loans competitively means that banks will sell the loans for

62

3 The Impact of Income Diversification on Chinese Banks …

which they possess no comparative advantage in financing. However, such restrictions offer non-interest activities a comparative advantage, as non-interest objectives have a low reserves requirement. Consequently, capital adequacy rules encourage banks to extend their sources of non-interest income. Based on the theory of Flannery (1989) showed that compared with the traditional activities, non-interest activities are less likely to be affected by capital restrictions. Hence the different capital restriction requirements would make banks within the three categories maintain different non-interest expansion strategies and might lead to different results in terms of the diversification effects. Finally, owing to the differences in the ownership structure, the three banking categories, i.e. G-SIBs D-SIBs and N-SIBs, have different internal governance mechanisms and face varying degrees of government intervention, all of which contribute to creating different diversification effects. For example, all four G-SIBs are stateowned banks, while D-SIBs are national joint-stock banks and N-SIBs are city and rural commercial banks controlled by local governments. For the reasons stated above, in the following empirical chapters the main sample includes 40 Chinese commercial banks, occupying 79% of total assets of the Chinese banking industry. The banks are divided into three groups: global systemically important banks (G-SIBs), domestic systemically important banks (D-SIBs), and other banks, which are deemed as systemically important (N-SIBs). Drawing from Basel III and the China Banking Regulatory Commission (CBRC), Table 3.1 lists the banks in the sample.

3.3.3 Research Methodology Existing studies, such as Stiroh and Rumble (2006), Baele et al. (2007), and Gamra and Plihon (2011), widely adopt pooled OLS estimation. Some studies find that income diversification would also impact on banking strategies, and that more risky banks are more likely to diversify (Acharya et al. 2006). In 1995, Berger and Ofek identified a diversification discount without controlling for endogeneity. However, subsequent papers, such as Campa and Kedia (2002) and Villalonga (2004), find that with consideration of the endogeneity problem they get an inverse result; that is, the result becomes positive with the same methodology. More importantly, studies such as Acharya et al. (2006), Stiroh and Rumble (2006) and Baele et al. (2007) also maintain that it is necessary to control for endogeneity, because diversification strategies are correlated with the banks’ business opportunities. Furthermore, a number of bank-specific characteristics, such as omitted management strategy variables (Gurbuz et al. 2013) and sensitivity of bank risk level to macroeconomic shocks (Berger et al. 2000), might lead to bias in the estimation and thus increase potential endogeneity concerns (Nisar et al. 2018). To address this endogeneity problem, Arellano and Bond (1991) propose taking the first difference in order to eliminate the fixed effect and using of difference GMM (DIF-GMM) for the model estimation. However, the proposed DIF-GMM

3.3 Variables, Data and Methodology

63

Table 3.1 List of the banks in three categories No.

Global systemically important Total assets (million banks USD)

Number of employees (thousand)

1

Industrial and Commercial Bank of China

4,006,242

453

2

China Construction Bank

3,397,688

353

3

Agricultural Bank of China

3,233,212

487

4

Bank of China

2,989,653

311

Domestic systemically important banks 5

Bank of Communications

1,388,023

91

6

Industrial Bank

985,448

62

7

China Merchants Bank

967,141

73

8

Shanghai Pudong Development Bank

942,509

54

9

China Minsheng Bank

906,396

58

10

China CITIC Bank

871,935

57

11

China Everbright Bank

627,840

44

12

Hua Xia Bank

385,301

43

13

Ping An Bank

199,682

14

Non-systemically important banks 14

Bank of Beijing

357,793

15

15

China Guangfa Bank

318,342

37

16

Bank of Shanghai

277,623

10

17

China Zheshang Bank

236,002

13

18

Bank of Nanjing

175,251

9

19

Hengfeng Bank

173,893

11

20

Bank of Ningbo

158,493

12

21

Shengjing Bank

158,274

5

22

China Bohai Bank

153,966

7

23

Huishang Bank

139,459

10

24

Chongqing Rural Commercial Bank

139,102

16

25

Bank of Hangzhou

127,978

7

26

Shanghai Rural Commercial Bank

123,174

6

27

Chengdu Rural Commercial Bank

108,355

8

28

Bank of Tianjin

107,794

7

29

Bank of Harbin

86,654

7

30

Bank of Changsha

72,262

6 (continued)

64

3 The Impact of Income Diversification on Chinese Banks …

Table 3.1 (continued) No.

Global systemically important Total assets (million banks USD)

Number of employees (thousand)

31

Bank of Guangzhou

67,595

4

32

Bank of Zhengzhou

66,931

4

33

Bank of Chengdu

66,733

6

34

Bank of Chongqing

64,925

4

35

Bank of Dalian

58,659

5

36

Bank of Hebei

51,717

5

37

Bank of Kunlun

48,763

5

38

Bank of Qingdao

47,035

4

39

Guangdong Shunde Rural Commercial bank

45,743

4

40

Bank of Dongguan

40,126

4

Source Author’s creation according to China Banking Regulatory Commission

method has the problem of weak instruments, and it would exacerbate measurement error biases (Griliches and Hausman 1986). Thus, when the instruments are weakly correlated with the explanatory variables, the DIF-GMM estimation would become close to the OLS-biased estimation. Developed on the DIF-GMM method, Blundell and Bond (1998) introduce the system GMM (SYS-GMM) approach, which combines both level value and differentiation value in order to reduce potential biases. SYS-GMM adds the exogenous difference of lagged instrument variables to the level equation. Selection of proper instrumental variables can solve the endogeneity problem and allow effective estimation of the panel data. We follow this approach in adopting the two-step robust standard error estimation for the dynamic model. GMM was originally proposed by Arellano and Bover (1995) and Blundell and Bond (1998). It entails no particular distribution assumptions and the random error terms are allowed to have heteroscedasticity and sequence correlation (Back and Brown 1993; Harvey and Zhou 1993). System GMM improves upon difference GMM, which suffered from weak instrument problems, by building a system of two equations - the original equation and the transformed one. In the first place, a normal dynamic regression model can be written as: yi,t = αyi,t−1 + β  X i,t + μi + ∈i,t

(3.7)

In this book, yi,t refers to the profitability of bank i in the period t. yi,t−1 is the lag term for profitability indicator. μi is fixed effect and ∈i,t .is error term. X i,t . covers bank’s diversification level and other control variables, including the ratios of net interest income to total earning assets, of loans to total as, and of non-interest expenses to total expenses. Other potential impact factors, such as size, regulatory

3.3 Variables, Data and Methodology

65

differences and the extent of moral hazard, as discussed above, can be evaluated over the three sub-groups. For Eq. 3.7, the usage of OLS will result in biased and inconsistent estimation because the lagged dependent variable is correlated with the residuals. In order to remove the bias, Holtz-Eakin et al. (1988) and Arellano and Bond (1991) proposed the first-difference transformation of (3.7) as follows:     yi.t − yi.t−1 = α(yi.t−1 − yi.t−2 ) + β  X i,t − X i,t−1 + ∈i,t − ∈i,t−1 .

(3.8)

Next, in Eq. 3.8, the control variables might not be strictly exogenous; rather, they might be related with the new error term, thus introducing pential endogeneity. Arellano and Bond (1991) introduced the lagged levels of the explanatory variables as instruments under the assumptions that the error term, ∈i,t . is not serially correlated and that the explanatory variables are weakly exogenous. This dynamic panel estimator is referred to as difference GMM, where its moment conditions are:    E yi,t−l ∈i,t − ∈i,t−1 = 0 for l ≥ 2; t = 3, . . . , T.

(3.9)

   E X i,t−l ∈i,t − ∈i,t−1 = 0 for l ≥ 2; t = 3, . . . , T

(3.10)

However, as the difference GMM might suffer from the weak instruments problem, especially under the condition of small sample size, system GMM is used to augment the difference estimator by estimating simultaneously in both differences and levels, with the two equations being distinctly instrumented. The additional moment conditions for the regression in level are:    E (yi.t−l − yi.t−l−1 ) μi + ∈i,t = 0 for l = 1

(3.11)

   E (X i.t−l − X i.t−l−1 ) μi,t + ∈i,t = 0 for l = 1

(3.12)

In order to test the reliability of estimation, this book employs two diagnostic tests commonly used with system GMM. First, the Arellano and Bond (1991) test is used to check for autocorrelation in the residuals AR (1) and AR (2). Then, as the effectiveness of SYS-GMM is largely dependent on whether the instrumental variables are exogenous, in order to avoid the over-identifying restrictions (Chiorazzo et al. 2008), this study also uses the Sargan test (Sargan 1958) to test the jointinspection of  −1 ˆ instrumental variables. Sargan statistics can be described as N−1 Z Eˆ Z Z Z E, which performs as a kind of Wald test, measuring the asymptotic chi-square distribution with degrees of freedom equal to the difference between the number of moments and parameters. Specifically, in log form the model specification is:

66

3 The Impact of Income Diversification on Chinese Banks …

PROi,t = αo + β1 PROi,t−1 + β2 HHIi,t + β3 NIMi,t + β4 LTAi,t + β5 NONi,t + εi,t (3.13) where empirical results of Eq. 3.13 are reported in Table 3.3 and Table 3.4 for whole Chinese banking industry and in Tables 3.6, 3.8, 3.10, 3.11, 3.12 and 3.13 for three Chinese banking groups (G-SIBs, D-SIBs and N-SIBs). Subscript i indicates the ith bank; t is the time period; PRO is profitability (ROA, ROE); the risk-adjusted profitability is represented by RAROA or RAROE; HHI is the Herfindahl-Hirschman Index and the shares of three non-interest components over total income; NIM, LTA and NON are the control variables, namely the ratio of net interest income to total earning assets, the loans to total assets and the ratio of non-interest expenses to total expenses respectively.

3.4 Empirical Research and Analysis 3.4.1 Income Diversification and Performance: Whole Sample Table 3.2 reports the descriptive statistics for our pooled sample. Banks’ diversification level is measured by HHI, which ranges from 0.418 to 40.600. The mean value of HHI is 14.690, which is far lower than that of banks from mature markets (Elsas et al. 2010) and other emerging markets (Sanya and Wolfe 2011). This result indicates that the level of diversification in Chinese banks overall is low and that they have a high concentration of interest earning activities. In addition, bank performance Table 3.2 Descriptive statistics for Chinese banks from 2005 to 2016 Mean

Median

SD

HHI

14.690

13.470

7.751

0.418

40.600

3.119

0.677

ROA

0.983

1.050

0.332

−0.201

2.227

3.926

−0.679

ROE

17.430

41.780

11.780

−1.416

2.733

0.376

17.50

6.890

Min

−27.92

ROROA

4.196

4.023

2.095

−0.509

RAROE

4.228

4.012

2.415

−2.088

NIM

2.769

2.775

0.582

0.701

LTA

45.950

47.170

9.389

14.380

NON

0.892

0.909

0.254

0.0330

Max

9.922 14.23

Kurtosis

Skewness

4.367

0.785

4.544

3.627

−0.321

69.770

3.105

−0.368

2.173

4.713

0.187

Variable definition: HHI: income diversification using Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income; ROA: return on assets; ROE: return on equities; RAROA: risk-adjusted return on assets; RAROE: risk-adjusted return on equity; NIM: total interest income/total interest expenses; LTA: loans/total assets; NON: non-interest expenses/total assets

3.4 Empirical Research and Analysis

67

widely varies within the Chinese banking sector. The alternative measures of bank performance, i.e., ROA, ROE, RAROA and RAROE, have a mean value of 0.983, 17.430, 4.196 and 4.228, respectively. Table 3.3 reports the results from estimating the dynamic panel models. Accoridng to Table 3.3, first, evidence from the two-step SYS-GMM regression shows that lagged income diversification (HHI) is positively correlated with present bank performance. This result indicates an accelerator effect from diversification, where past performance has a positive effect on future performance. Next, we find that diversification is negatively associated with profitability (ROA, ROE), indicating a performance decrease from income diversification. To examine the robustness of the result, we also investigate the likely impact on the risk-adjusted performance indicators of both RAROA and RAROE. Overall, for the whole sample, the results show negative effects of diversification on banks’ both performance and risk-adjusted performance. Table 3.3 Income diversification and profitability for Chinese banks, 2005–2016 ROA

ROE

RAROA

RAROE

0.437***

0.529***

0.577***

0.808***

(0.021)

(0.022)

(0.024)

(0.042)

−0.002**

−0.102***

−0.047***

−0.019*

(0.001)

(0.016)

(0.003)

(0.008)

NIM

0.145***

0.069

0.336***

0.083

(0.004)

(0.227)

(0.080)

(0.047)

LTA

−0.002***

−0.067*

−0.006

−0.007

(0.001)

(0.027)

(0.006)

(0.008)

0.120***

4.356***

0.587**

0.340

(0.011)

(0.777)

(0.201)

(0.196)

Constant

0.122*

8.625***

0.253

1.321*

(2.190)

(7.190)

(1.410)

(2.250)

F-test

0.000

0.000

0.000

0.000

Sargan test

1.000

1.000

0.512

0.999

It-1 HHI

NON

AR(2)

0.133

0.159

0.900

0.119

Observations

419

415

419

415

This table reports the two-step SYS-GMM dynamic panel estimation results. Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). HHI indicates income diversification by using the Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

68

3 The Impact of Income Diversification on Chinese Banks …

In general, our study echoes the work of Lepetit et al. (2008) and Mercieca et al. (2007), who use mature market data. According to Köhler (2015), the performance discount in mature banking markets mainly originates from over-diversification. However, income diversification in the Chinese banking sector overall is quite low. As suggested by Wagner (2010), a non-linear relationship exists between banks’ diversification and performance; at both the lower and the higher levels of diversification, banks are unable to optimize their performance. Hence, it is plausible that the performance decrease for Chinese banks is driven by under-diversification, where non-interest activities incur high initial costs in the early stages of diversification. Given the average low level of income diversification in the Chinese banking sector, banks require several years to absorb these initial costs; hence, we see an overall diversification discount in this market. In addition, at the stage of low diversification, few managers have appropriate skills to engage in non-interest business. Because non-interest activities have a higher level of relevance among different products than traditional activities, reliance on unskilled staff leads to losses and to a significant reduction in profitability. On the other hand, this implies high initial costs of training specialized workers, which would cause a significant reduction in profitability. Owing to the high level of fixed expenses of non-interest activities, the costs of non-interest income significantly increase and thus net profits become negative, thus causing a reduction in diversification benefits. It is also plausible that the Chinese banking sector is under-regulated. Managers in banks with abundant cash flows would invest in low profit projects, which would increase the moral hazard problem and thus lead to an increase in capital costs (Easley and O’hara 2004) and inefficient resource allocation (Fisher et al. 2002). Consequently, the value decrease from diversification would be accelerated. The table also reports several diagnostic test results. The results presented in the last four rows of Table 3.3 show that the F-statistics for all models with four performance indicators are significant. To check for autocorrelation, we use the ArellanoBond test for autocorrelation serial correlation (AR2). Second-order autocorrelation is statistically nonsignificant. We employ the Sargan test to examine whether our models include effective instruments, and all P-values of the Sargan tests are above the 10% significance level, which indicates that the instruments satisfy the orthogonality conditions required for their employment. It is conceivable that individual components of non-interest business may perform differently than the overall non-interest activities. To find further evidence for diversification effects across different components of non-interest income, we divide the non-interest income into three categories, namely, fee and commissions, trading, and other income, and the results are reported in Table 3.4. As can be seen from the table, the results show that sub-businesses under non-interest activities exert different effects on banks’ performance level. Fee and commissions and other activities show a negative effect among our four performance indicators, while trading activities have positive coefficients. This indicates that the diversification discount for the Chinese banking sector is generated mainly from commissions and other non-interest activities, while trading activities would lead to improvements in profitability.

Constant

NON

LTA

NIM

OTH

TRAD

COM

It-1

(0.023)

(0.028)

(0.019)

0.578***

RAROA (0.037)

0.823***

RAROE

8.226***

(1.185)

0.138***

(0.033)

(0.615)

(0.010)

(0.018)

(0.001)

4.373***

−0.057**

−0.001

0.105***

0.068

(0.182)

0.131***

(0.039)

(0.007)

(0.001)

(0.440)

0.994*

(0.239)

0.540*

(0.007)

−0.004

(0.095)

0.386***

(0.009)

(0.406)

0.669

(0.212)

0.370

(0.007)

−0.005

(0.070)

0.072

(0.009)

−0.005*** −0.180*** −0.074*** −0.037***

0.523***

ROE

0.498***

ROA

ROE

(0.033)

0.047

(0.018)

(0.038)

0.128***

(0.007)

0.021**

(0.013)

0.824***

RAROA

(0.024)

(1.015)

10.807***

(0.732) (0.136)

0.160*** (0.009)

(0.262)

0.104

(0.278)

(0.086)

0.751***

(0.042)

(0.146)

1.887***

(0.006)

(0.199)

(0.047)

0.427***

(0.038)

0.265***

(0.001)

(0.721)

7.680***

(0.559)

7.859***

(0.014)

(0.531)

2.174***

(0.009)

(0.338)

(continued)

(0.430)

−0.915** −0.528

(0.293)

0.754*

(0.007)

0.016

(0.224)

−0.940***

−0.132***

(0.028)

1.136***

RAROE

(0.023)

(0.040)

0.618***

RAROA

−0.099*** −0.965*** −0.092*

(0.028)

0.382***

ROE

(0.008)

(0.023)

0.346***

ROA

−0.022*** −0.007*** −0.061*** −0.002

(0.030)

−0.070*

(0.003)

0.024***

(0.013)

0.705***

RAROE

−0.935*** 0.748***

(0.052)

1.349***

(0.003)

−0.271*** 0.006

(0.207)

0.121*** 13.525***

(0.001)

−0.002*

(0.010)

0.173*** 0.200

0.105*** (0.027)

0.006**

(0.014)

(0.002)

(0.019)

0.461*** 0.390***

ROA

Table 3.4 Results for three components of non-interest income and bank performance for Chinese banks, 2005–2016

3.4 Empirical Research and Analysis 69

1.000

0.201

Sargan test

AR(2)

415

0.158

1.000

0.000

ROE

419

0.957

0.394

0.000

RAROA

415

0.111

1.000

0.000

RAROE

410

0.250

0.982

0.000

ROA

406

0.253

0.740

0.000

ROE

410

0.191

0.182

0.000

RAROA

RAROE

406

0.123

0.676

0.000

ROA

417

0.143

0.780

0.000

ROE

413

0.159

0.902

0.000

RAROA

417

0.744

0.254

0.000

RAROE

413

0.144

1.000

0.000

This table reports the two-step SYS-GMM dynamic panel estimation results with robust errors. Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). It-1 refers to the lagged dependent variables by one period. COM is the ratio of net fee and commission incomes to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

Observations 419

0.000

F-test

ROA

Table 3.4 (continued)

70 3 The Impact of Income Diversification on Chinese Banks …

3.4 Empirical Research and Analysis

71

Table 3.5 Descriptive statistics for Chinese G-SIBs, 2005–2016 HHI

Mean

Median

23.560

23.570

SD 5.696

Min 11.650

Max 40.190

Kurtosis

Skewness

3.428

0.138

ROA

1.104

1.168

0.301

0.024

1.475

6.836

−1.705

ROE

15.100

17.960

10.660

−27.920

23.430

11.010

−2.891

ROROA

5.705

5.595

2.634

0.063

9.157

1.899

−0.328

RAROE

4.532

6.326

3.736

−2.088

9.422

1.364

−0.062

NIM

2.804

2.845

0.504

1.049

3.626

4.983

−1.032

LTA

50.610

51.090

3.872

42.980

2.251

0.072

NON

1.002

0.999

0.168

0.628

2.978

−0.089

58.96 1.396

Variable definition: HHI (%): income diversification using the Herfindahl–Hirschman Index with the components of interest income and three activities under non-interest income; ROA: return on assets; ROE: return on equities; RAROA: risk-adjusted return on assets; RAROE: risk-adjusted return on equity; NIM (%): total interest income/total interest expenses. LTA (%): loans/total assets, NON (%): non-interest expenses/total assets

3.4.2 Income Diversification and Performance Across Chinese Banking Groups 3.4.2.1

China’s Global Systemically Important Banks (G-SIBs)

Having examined the performance of the Chinese banking sector for the whole sample, we next examine sub-samples of Chinese banks, categorized according to their systemic importance. Table 3.5 presents summary statistics for the group of China’s global systemically important banks. From the table, the mean of the HHI for G-SIBs is 23.560, which is significantly higher than that for the whole sample (14.690). The minimum value of the HHI as a measure for the level of diversification is 11.650, while the maximum is 40.190. Meanwhile, four performance measurements—ROA, ROE, RAROA and RAROE—have a mean value of 1.104, 15.100, 5.705 and 4.532, respectively. The mean of ratio of non-interest income to total assets (NIM) is 2.804%, with a range from 1.049 to 3.626%. The mean loan to assets (LTA) is 50.610%, with a minimum of 42.980% and a maximum value of 58.960%. The mean non-interest expenses over total assets (NON) is 1.002%, with a minimum of 0.628% and a maximum value of 1.396%. There is no significant skewness in the sample. The values of skewness are within the acceptable and expected ranges, indicating that there is no evidence of the data being skewed toward either extreme. Next, Table 3.6 reports the results from estimating the association between the income diversification of G-SIBs and their performance during the period 2005– 2016.2 The results indicate that diversification is an important determinant of banks’ 2 As the small sample size might result in an incidental parameters problem, this book also applies a

robustness test by using dummy variables to the catalogue of the three sub-groups. The robustness test results are reported in the Appendix.

72

3 The Impact of Income Diversification on Chinese Banks …

Table 3.6 Income diversification and profitability for Chinese G-SIBs, 2005–2016 ROA

ROE

RAROA

RAROE

It-1

0.764***

0.347

0.980***

0.969***

(0.075)

(0.211)

(0.071)

(0.042)

HHI

0.013***

0.539***

0.082*

0.021*

(0.003)

(0.105)

(0.041)

(0.010)

0.0583

5.716***

−0.0686

−0.491**

(0.031)

(1.646)

(0.175)

(0.162)

LTA

−0.010

−0.441**

−0.093

−0.093*

(0.009)

(0.160)

(0.053)

(0.041)

NON

0.221***

0.221

1.758**

1.548*

(0.054)

(1.622)

(0.541)

(0.717)

0.102

3.975

1.473

4.287*

(0.270)

(0.420)

(0.450)

(1.980)

F-test

0.000

0.001

0.000

0.000

Sargan test

1.000

1.000

0.998

1.000

AR(2)

0.102

0.260

0.107

0.132

Observations

44

42

43

42

NIM

Constant

Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). HHI is income diversification by using the Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income. NIM is total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

performance. From the SYS-GMM estimations, the coefficients of the HHI are significantly and positively associated with all the four performance indicators. The evidence thus shows that diversification has a positive impact on the largest banks in China, i.e., Chinese G-SIBs. This result indicates that Chinese G-SIBs can benefit from diversifying into non-traditional businesses. Consequently, the higher reliance on non-interest income could make G-SIBs more profitable. The positive results of the performance effect of diversification are significant in that these G-SIBs are of critical importance to the Chinese banking system. The aggregate assets of these G-SIBs account for 49% of the total assets of all Chinese banks, and hence, they are the key player in the construct of China’s banking industry. Their success bolsters the stability of the Chinese banking system and also gives a great boost to other banks shifting to non-traditional business. The main source of their success seems to lie in the fact that these big banks are well positioned to exploit the economy of scope resulting from the diversification. As suggested by Gurbuz et al. (2013), large-sized banks generally have better information technology, human capital management and risk management. Therefore, such

3.4 Empirical Research and Analysis

73

business expansion could improve the overall productivity and cause technology spillover within the banking system (Canals 1994; Acharya et al. 2006; Mercieca et al. 2007). In addition, for these established banks, the initial cost incurred from shifting to non-interest business, including building, IT facilities, business reputation and advertisement, can also be largely shared with the traditional business. The learning-by-doing effect may also have worked for Chinese G-SIBs. Understandably, the initial stages of bank diversification would incur sizable operating losses owing to, say, unexperienced personnel who are unfamiliar with new business lines. However, with expansion of the diversification, such a disadvantage would be offset by the accumulation of more experienced staff or by the learning-by-doing effect (Gamra and Plihon 2011). With this effect, banking institutions can achieve performance improvement through practice, self-perfection and minor innovations. Consequently, they can progress to reap diversification benefits as long as they have taken care of diversifying according to their specific characteristics, competences and risk levels. It is reasonable to infer that this process may have also occurred with Chinese G-SIBs. The regulatory difference is another key factor in Chinese G-SIBs’ performance gain from diversification. China’s Banking Regulatory Commission has implemented different levels of financial restrictions on the three groups, where the lower boundary of the core tier-one capital requirement is higher for G-SIBs than for the other two, at 8%, while the lower boundary for the capital adequacy ratio is 11.5%. Such tight restrictions are an important factor that explains why G-SIBs seek higher levels of diversification. Non-interest activities have the comparative advantage of high reserve requirements. Consequently, the capital adequacy rules act as an incentive for banks to extend their business to earn non-interest income. Further, the stricter capital restriction of banks’ activities would also increase banks’ effectiveness (Agoraki et al. 2011) and change banks’ risk preference (Flannery 1989). Consequently, a more risk-averse, effective resources-allocation strategy and strict cost management could have led to income diversification benefits in Chinese G-SIBs. The results of diagnostic tests are reported in the lower panel of Table 3.6. All test results are satisfactory across all model specifications. The P-values of the F tests for the four models are close to zero, indicating the joint significance of our regressors. Regarding the efficiency of the GMM estimation, the results of the Sargan test are nonsignificant; hence, our instruments are appropriately orthogonal to the error. In addition, the coefficient of the AR (2) tests for the second-order serial correlation are nonsignificant at the 1% significance level.

3.4.2.2

China’s Domestic Systemically Important Banks

We now move to examine the performance effect of diversification for China’s domestic systemically important banks (D-SIBs). Table 3.7 presents the summary statistics. Compared with G-SIBs, this group has a considerably lower level of diversification: the mean value of the HHI for D-SIBs is 17.230, while the corresponding figure for G-SIBs is 23.560. Regarding the four performance indicators, there is no

74

3 The Impact of Income Diversification on Chinese Banks …

Table 3.7 Descriptive statistics for Chinese D-SIBs, 2005–2016 HHI

Mean

Median

SD

17.230

15.610

7.977

Min 4.789

Max

Kurtosis

40.600

2.685

Skewness 0.555

ROA

0.952

0.999

0.280

0.133

1.460

3.128

−0.673

ROE

19.020

18.380

5.245

4.176

41.130

6.018

0.837

ROROA

4.208

4.263

1.602

0.431

7.925

2.732

0.107

RAROE

4.976

5.171

1.707

0.586

8.038

2.424

−0.285

NIM

2.764

2.797

0.387

1.733

3.847

3.192

−0.336

LTA

51.950

51.690

6.958

33.580

67.360

3.225

−0.091

NON

0.951

0.959

0.194

0.502

1.439

2.860

−0.168

Variable definitions: HHI (%): income diversification using the Herfindahl–Hirschman Index with the components of interest income and three activities under non-interest income; ROA: return on assets; ROE: return on equities; RAROA: risk-adjusted return on assets; RAROE: risk-adjusted return on equity; NIM (%): total interest income/total interest expenses. LTA (%): loans/total assets, NON (%): non-interest expenses/total assets

significant difference between D-SIBs and G-SIBs. The mean value of ROA, ROE, RAROA and RAROE is 0.952, 19.020, 4.208 and 4.976, respectively. Moreover, the mean non-interest expenses over total assets (NON) are lower than those for G-SIBs, implying that the input for non-interest activities by G-SIBs, such as professional training and initial investment, is relatively lower than that of G-SIBs. To ensure the robustness, we estimate two different kinds of models, one for profitability (ROA, ROE) and the other for risk-adjusted performance (RAROA, RAROE), using the two-step SYS-GMM estimator with robust standard errors procedures. The results for D-SIBs are displayed in Table 3.8. Similar to the case of G-SIBs, the HHI index of D-SIBs is positively correlated with both sets of performance indicators. D-SIBs can gain a performance improvement from income diversification. However, compared with those for G-SIBs, the coefficients for D-SIBs are relatively small, and all are nonsignificant. The results may be explained by the fact that banks in this group have a smaller size than the G-SIBs, and hence, the improvement from economy of scope might not be sufficiently large to offset the performance. For these relatively small banks, the high initial diversification cost and staff’s lack of experience might lead to them showing no significant performance gains from diversification. We also report the diagnostic test results. The F-test yields a significant P-value at the 5% level, indicating that variables in the models are not jointly nonsignificant. Regarding the efficiency of GMM estimation, as all coefficients of the AR (2) and the Sargan tests are nonsignificant, we can conclude that the instruments used are not correlated with the residuals, and there is no problem of autocorrelation. Thus, the models are reasonable and statistically acceptable.

3.4 Empirical Research and Analysis

75

Table 3.8 Income diversification and profitability for Chinese D-SIBs, 2005–2016 ROA

ROE

RAROA

RAROE

It-1

0.523***

0.999

0.646***

0.743***

(0.056)

(1.225)

(0.088)

(0.082)

HHI

0.001

0.002

0.039

0.065

(0.002)

(0.677)

(0.033)

(0.046)

0.249*

−1.548

0.110

−1.008

(0.090)

(8.686)

(0.532)

(0.781)

LTA

−0.003

0.266

0.089

0.150*

(0.002)

(1.017)

(0.050)

(0.074)

NON

0.003

2.075

2.112

3.636

(0.115)

(18.300)

(1.245)

(1.870)

−0.083

−11.560

−5.894

−8.214

(−0.320)

(−0.160)

(−1.520)

(−1.460)

F-test

0.000

0.000

0.000

0.000

Sargan test

0.495

1.000

0.992

0.327

AR(2)

0.573

0.434

0.568

0.172

Observations

99

97

97

97

NIM

Constant

Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). HHI is income diversification by using the Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income. NIM is total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

3.4.2.3

Other Chinese Banks (N-SIBs)

Table 3.9 reports the summary statistics for other Chinese banks (N-SIBs). Compared with banks in the G-SIBs and D-SIBs groups, N-SIBs have the lowest mean of diversification, at 12.390, which is much lower than that of the G-SIBs (23.560). Interestingly, compared with G-SIBs and D-SIBs, N-SIBs have a similar level of mean profitability (ROA, ROE). However, concerning risk-adjusted performance, both RAROA and RAROE variables are significantly lower for N-SIBs than for the other two groups. This result implies that Chinese N-SIBs have poor risk management. In addition to banks’ income diversification, we control for several other characteristics that might affect bank performance. The mean net interest margin (NIM) is 2.765%, with a minimum of 0.701% and a maximum value of 4.544%. The mean loan to assets (LTA) is 43.090%, with a minimum of 14.380% and a maximum value of 69.770%. Meanwhile, N-SIBs have the lowest ratio of non-interest expenses to total assets (NON), at 0.854% compared with 1.002% for G-SIBs and 0.951% for D-SIBs. This

76

3 The Impact of Income Diversification on Chinese Banks …

Table 3.9 Descriptive statistics for Chinese N-SIBs, 2005–2016 HHI

Mean

Median

SD

12.390

11.940

6.591

1.040

0.350

−0.201

2.227

3.833

−0.587

6.550

−15.700

41.780

6.105

−0.399 0.418

ROA

0.975

ROE

17.240

17.03

Min 0.418

Max

Kurtosis

37.760

4.416

Skewness 0.926

ROROA

3.955

3.843

2.060

−0.509

9.922

2.863

RAROE

3.921

3.639

2.312

−1.452

14.230

6.783

1.475

NIM

2.765

2.752

0.649

0.701

4.544

3.158

−0.247

LTA

43.090

42.960

9.468

69.770

2.991

−0.107

NON

0.854

0.867

0.275

2.173

4.966

0.471

14.38 0.033

Variable definitions: HHI (%): income diversification using the Herfindahl–Hirschman Index with the components of interest income and three activities under non-interest income; ROA: return on assets; ROE: return on equities; RAROA: risk-adjusted return on assets; RAROE: risk-adjusted return on equity; and NIM (%): total interest income/total interest expenses. LTA (%): loans/total assets, NON (%): non-interest expenses/total assets

result indicates that staff training costs and potential losses from non-professional operation are lower for N-SIBs, leading to a higher probability of operational loss. Then, in Table 3.10, we report the regression results for the effect of diversification on four bank performance variables when adopting the two-step SYS-GMM dynamic panel model. We can see that for N-SIBs, shifting from traditional banking towards non-interest activities significantly reduces their performance in terms of both profitability and risk-adjusted profitability, which means that for banks with low levels, diversification into non-traditional income will adversely affect their returns and risk. For Chinese N-SIBs, non-interest income has consistently accounted for only a small proportion of their total operational income. Positive impacts of diversification into non-interest activities on their profitability, if any, could hardly be sizable. Rather, owing to the necessary expenses, the cost of non-interest income could be higher than the possible gains, turning net performance into negative. Moreover, with a low level of diversification, these banks could hardly benefit from the learning-by-doing effect; instead, they lack a sufficient number of experienced workers to improve the rationality of operating decisions, which negatively affects their performance. These N-SIBs also suffer from financial deregulation. Compared with G-SIBs and D-SIBs, N-SIBs operate under lower financial restrictions. Consequently, managers have less incentive to seek better business opportunities through financial innovation, which would in turn diminish banks’ performance owing to problems such as moral hazard. Agoraki et al. (2011) maintain that a more relaxed capital restriction would reduce banks’ effectiveness. Consequently, an ineffective and less risk-averse resource allocation strategy and cost management under financial deregulation could lead to an income diversification discount. Size also matters here. The majority of Chinese N-SIBs are small-sized banks. Furthermore, they invest far fewer resources in new business lines. As shown in Table 3.8, the ratios of NON are rather diverse among G-SIBs, D-SIBs and N-SIBs.

3.4 Empirical Research and Analysis

77

Table 3.10 Income diversification and profitability for Chinese N-SIBs, 2005–2016 ROA

ROE

RAROA

RAROE

It-1

0.465***

0.612***

0.657***

0.939***

(0.032)

(0.048)

(0.085)

(0.025)

HHI

−0.003***

−0.131***

−0.022***

−0.037***

(0.001)

(0.027)

(0.007)

(0.010)

0.141***

0.0363

−0.131

−0.0981

(0.023)

(0.301)

(0.110)

(0.082)

LTA

−0.002

−0.088**

−0.025

−0.006

(0.002)

(0.032)

(0.014)

(0.007)

NON

0.063

3.745**

1.786**

0.528*

(0.042)

(1.285)

(0.561)

(0.247)

0.366***

11.510***

1.148

0.476

(5.310)

(6.940)

(1.740)

(0.800)

F-test

0.000***

0.000***

0.000***

0.000***

Sargan test

1.000

1.000

0.999

1.000

AR(2)

0.161

0.567

0.838

0.813

Observations

276

276

276

276

NIM

Constant

Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). HHI is income diversification by using the Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

While the average NON for the Chinese banking sector as a whole is 0.892, G-SIBs have the highest mean value (1.002), followed by D-SIBs (0.951) and N-SIBs scores (with only 0.854). Thus, N-SIBs invest fewer resources into non-interest activities. As they lack efficient resources for shifting to new financial products and the relevant experience needed to manage the new product mix, it is difficult for these small banks to exploit economies of scope since they have limited technical capacity and since they cannot provide a lower marginal cost for their financial product in order to offset the increase in fixed costs or inefficient risk controls (Mercieca et al. 2007). The last three rows in Table 3.10 present the diagnostic test results for the models. All models pass the F tests, and their construction is acceptable, with all variables not jointly nonsignificant. The results for both the AR (2) and Sargan tests are nonsignificant for all models, suggesting that we can conclude that the instruments used in the GMM models are reasonable and statistically acceptable.

78

3 The Impact of Income Diversification on Chinese Banks …

3.4.3 Effects of Diversification on Performance by Components of Non-interest Activities Following the arrangement of the previous section, we have also subdivided the noninterest income into three categories, and then studied whether the three types of diversification have different effects on different bank groups. The results presented in Table 3.11 indicate that the fee-based activities have different effects for G-SIBs than for the banking sector as a whole; that is, there is a significant and positive effect on the performance of G-SIBs, rather than a diversification discount. Turning to trading activities, we suggest that this business line will also improve bank performance among G-SIBs. The only negative effect for banks in this group is generated from other activities, which can be explained by the high leverage nature of those activities, and by a lack of skilled workers. The results in Tables 3.12 and 3.13 presents the effects of components of noninterest activities on bank performance for D- and N-SIBs. Table 3.12 shows that different types of non-interest business also have different impacts on the performance level of D-SIB banks. Specifically, fee and commission activities have a negative impact on bank performance. The results for transactions and other sub-activities show similar signs and directions to those for G-SIBs. However, as the level of income diversification is quite low relative to the level of fee collection activities, the expansion of such business has had a negative impact on bank performance. As can be seen from the table, among the three components of non-interest income, the proportion of commissions and other non-interest activities is relatively large, resulting in a decline in the performance of small banks, while various types of banks benefit from further participation in trading activities when their performance improves. This finding is consistent with the previous results for medium-sized DSIBs. We suggest that banks of different type and size receive different diversification effects with different components of non-interest activities. Trading activity can bring benefits in terms of performance improvement for the overall banking sector and for all three sub-groups. However, we see that the proportion of trading activities in banks’ income stream is still relatively low, especially for large and medium-sized banks, which are still paying more attention to fee-based activities. While such a diversification strategy can work well for G-SIBs, as they have already accumulated enough specialists and established a well-regulated risk management system, D-SIBs will eventually suffer a performance discount from too great an expansion of fee-based income. Therefore, rather than engaging blindly in diversification activities, banks’ managers should select the most appropriate diversification direction and strategy by considering their own situation and businesses scope. Furthermore, regulators should develop more detailed supervisory plans and guidance for banks’ diversification process by subdividing the components of non-interest business and taking into account the impact of different types of non-interest business on different types of banks.

Constant

NON

LTA

NIM

OTH

TRAD

COM

It-1

(0.148)

(0.007)

9.103

(9.448)

−0.112

(0.402)

(1.771)

4.803** (0.626)

5.294***

2.013*** (0.454)

(0.708)

(0.752)

0.288*** 7.786*** 2.623***

(0.029)

(0.053)

(0.026)

(0.244)

1.250***

(0.234)

0.647**

(0.004)

7.255*

(18.252) (3.313)

−3.068

(5.211)

2.536

(3.229)

−2.751** 0.432 (0.919)

(0.115)

−0.074

(1.205)

0.195

(0.216)

0.462*

(0.095)

−3.855 (8.401)

ROA

(0.027)

(1.001)

(0.167)

0.615***

(0.183)

0.614***

(0.004)

(8.949)

14.258

(1.517)

10.016***

(0.156)

−0.012** −0.158

(0.089)

(1.784)

4.913**

(0.457)

1.892***

(0.030)

−0.096**

(0.151)

(continued)

(0.048)

−0.195***

(1.161)

1.874

(0.056)

−0.035

(0.245)

−0.247** −3.971*** −0.707*** −0.975***

(0.081)

−0.195***

(0.044)

1.004***

RAROE

(0.048)

(0.041)

1.064***

RAROA

−0.023** −0.756*** −0.073**

(0.135)

0.826***

ROE

(0.007)

(0.039)

0.951*** 1.167***

RAROE

(0.081)

−0.166*

(0.443)

1.303**

(0.151)

0.259

(0.109)

0.990***

RAROA

(0.278)

(4.616)

−0.353* −0.154*** −0.132*** −0.032*** −0.053

(0.120)

−0.002

(0.091)

(1.130)

(0.355)

0.058

0.031

6.267

1.896 (2.296)

0.036*

(0.204)

0.400

ROE

(0.018)

(0.131)

−0.709*** −0.312**

(0.049)

0.144**

1.538***

ROA

(0.021)

−0.778*

(0.072)

(0.097)

(0.006)

(0.019)

0.804*** 0.197**

0.020**

(0.150)

(0.018)

RAROE

(0.050)

RAROA 0.954***

ROE

0.670*** 0.527*** 1.088***

ROA

Table 3.11 Results for three components of non-interest income and bank performance for G-SIBs, 2005–2016

3.4 Empirical Research and Analysis 79

0.999

0.116

Sargan test

AR(2)

42

0.317

1.000

0.000

ROE

44

0.154

0.999

0.000

RAROA

42

0.243

0.998

0.000

RAROE

43

0.125

1.000

0.000

ROA

41

0.158

1.000

0.000

ROE

43

0.111

0.996

0.087

RAROA

41

0.258

0.996

0.002

RAROE

44

0.162

1.000

0.000

ROA

ROE

42

1.000

1.000

0.000

44

0.102

1.000

0.000

RAROA

RAROE

42

0.249

0.779

0.000

This table reports the two-step SYS-GMM dynamic panel estimation results with robust errors. Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). It-1 refers to the lagged dependent variables by one period. COM is the ratio of net fee and commission incomes to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

Observations 44

0.000

F-test

ROA

Table 3.11 (continued)

80 3 The Impact of Income Diversification on Chinese Banks …

Constant

NON

LTA

NIM

OTH

TRAD

COM

It-1

(0.152)

RAROA

(0.039)

0.917***

RAROE (0.062)

2.162

(2.393)

8.693

(6.272)

0.075*

(0.034)

−0.034

(0.102)

(0.468)

−1.042* (1.734)

0.271

(0.968)

0.801

−0.061

(0.185)

(0.026)

(0.064)

(0.002)

0.023

(0.473)

−0.024

(0.027)

(0.007)

(0.116)

0.015*

(1.501)

−0.066

(0.020)

0.462***

(0.008)

0.002

2.599

(0.114)

0.068***

(0.001)

ROA

ROE

(0.887)

(0.436)

−0.401

(0.157)

−0.285

(0.003)

0.006*

(0.119)

RAROA

(0.342)

−0.372

(0.114)

0.304**

(0.083)

(0.028)

(0.040)

0.117**

(0.489)

−0.271

(0.191)

0.431*

(0.069)

(4.011)

−9.999*

(3.651) (1.491)

−2.676

(0.529)

(2.457)

−6.016*

(0.916)

10.621** 2.562*** 2.819**

(0.055)

RAROE

ROA

ROE (0.198)

0.218

RAROA (0.051)

0.897***

RAROE (0.069)

0.543***

(0.188)

−0.414*

(0.101)

0.193

(0.003)

0.007*

(0.064)

0.064

(0.018)

(7.283)

0.553

(3.588)

7.107*

(0.078)

0.074

(1.892)

1.928

(0.744)

(0.964)

−2.601**

(0.444)

0.866

(0.019)

0.043*

(0.234)

0.149

(0.098)

(continued)

(2.284)

2.934

(1.870)

5.398**

(0.035)

0.033

(0.915)

−2.418**

(0.414)

−0.058** −2.157** −0.325*** −0.952*

(0.048)

0.590*** 0.527*** 0.790***

0.202*** 0.057*

(2.123)

0.426*** 2.106

1.644

(0.022)

(0.108)

0.079

0.053*

(0.448)

0.672*** 0.163

−0.011*** −0.225* −0.055*** −0.063*

(0.063)

ROE

0.330*

0.810***

ROA

Table 3.12 Results for three components of non-interest income and bank performance for D-SIBs, 2005–2016

3.4 Empirical Research and Analysis 81

1.000

0.849

Sargan test

AR(2)

ROE

97

0.110

0.334

0.000

RAROA

99

0.271

1.000

0.000

RAROE

97

0.134

0.923

0.000

97

0.772

0.939

0.000

ROA

ROE

95

0.488

0.656

0.000

RAROA

95

0.487

0.861

0.000

95

0.140

0.967

0.000

RAROE

99

0.932

1.000

0.000

ROA

97

0.181

1.000

0.000

ROE

99

0.351

1.000

0.000

RAROA

93

0.151

0.978

0.000

RAROE

This table reports the two-step SYS-GMM dynamic panel estimation results with robust errors. Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). It-1 refers to the lagged dependent variables by one period. COM is the ratio of net fee and commission incomes to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

Observations 99

0.000

F-test

ROA

Table 3.12 (continued)

82 3 The Impact of Income Diversification on Chinese Banks …

Constant

NON

LTA

NIM

OTH

TRAD

COM

It-1

RAROE

(2.597)

(0.099)

(2.234)

9.772***

0.386***

(0.065)

(0.053)

4.932*

(0.002)

0.040

(0.551)

0.779

(0.273)

0.790**

(0.007)

0.003

(0.083)

(0.494)

−0.106*

(0.020)

0.182*

(0.018)

−0.094

−0.004*

0.140***

(0.070)

(0.318)

0.857**

(0.266)

0.590*

(0.004)

−0.003

(0.057)

−0.083

(0.017)

(0.093)

(0.004)

(0.085) −0.073***

(0.031)

RAROA

0.657*** 0.832***

−0.013** −0.254*** −0.037*

(0.062)

ROE

0.555***

0.440***

ROA

ROE

RAROA

RAROE

(0.092)

0.122

(0.054)

0.018

(0.002)

−0.002

(0.016)

(2.953)

6.228*

(1.637)

0.661

(0.045)

−0.021

(0.671)

0.190*** 1.427* (0.111) (0.005)

(0.521)

(0.289)

(0.277) −0.212

(0.298) −1.145*

1.072*** 1.392***

(0.009)

ROA

ROE

RAROA

RAROE

(0.024)

(0.038)

0.210***

(0.047)

0.553***

(0.001)

−0.001

(0.075)

(0.889)

−1.407

(0.992)

6.752***

(0.020)

0.137***

(0.601)

(0.273)

−0.218

(0.463)

2.003***

(0.008)

0.052***

(0.305)

(continued)

(0.544)

−0.143

(0.442)

1.200**

(0.009)

0.052***

(0.193)

−2.411*** −1.203*** −1.164***

(0.138)

(0.064)

(0.078)

1.106***

−0.222*** −0.330***

(0.053)

0.986***

(0.007)

(0.027)

0.845***

−0.022** −0.418**

(0.029)

0.594***

−0.496*** −0.059*

(0.022)

0.135***

(0.025)

0.034*** 0.012*

(0.190)

−0.236

(0.052)

(0.083)

(0.003)

(0.063)

0.159**

(0.017)

0.010*** 0.180*

(0.051)

0.411*** 0.478*** 0.822*** 0.940***

ROA

Table 3.13 Results for three components of non-interest income and bank performance for N-SIBs, 2005–2016

3.4 Empirical Research and Analysis 83

1.000

0.163

Sargan test

AR(2)

ROE

276

0.541

1.000

0.000

RAROA

276

0.741

0.992

0.000

RAROE

276

0.799

1.000

0.000

ROA

270

0.220

0.989

0.000

270

0.337

0.652

0.000

ROE

270

0.258

0.971

0.000

RAROA

270

0.384

1.000

0.000

RAROE

274

0.671

0.997

0.000

ROA

274

0.422

1.000

0.000

ROE

274

0.371

1.000

0.000

RAROA

274

0.593

1.000

0.000

RAROE

This table reports the two-step SYS-GMM dynamic panel estimation results with robust errors. Our dependent variables are return on assets (ROA), return on equities (ROE), risk-adjusted return on assets (RAROA), and risk-adjusted return on equity (RAROE). It-1 refers to the lagged dependent variables by one period. COM is the ratio of net fee and commission incomes to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

Observations 276

0.000

F-test

ROA

Table 3.13 (continued)

84 3 The Impact of Income Diversification on Chinese Banks …

3.5 Conclusion

85

3.5 Conclusion This chapter examines to what extent income diversification can affect performance in the Chinese banking industry. Employing a dynamic SYS-GMM panel data model to evaluate the performance effects of income diversification, this chapter finds existence of a diversification discount in the Chinese banking sector as a whole, suggesting that a shift from traditional banking business to mixed business lines negatively affects bank performance. However, structurally, the results are rather diverse. After separating the sample banks into three groups, we find that the largest Chinese banks, China’s global systemically important banks or G-SIBs, can improve their performance by using diversification. The next group, the domestic systemically important banks or DSIBs, shows a nonsignificant performance response to the shifting to mixed business lines. The most important under-performer is China’s non-systemically important banks or N-SIBs. The key factor that drives the performance differences lies in the banks’ capability to reap the benefits of diversification through the learning-by-doing process. Other factors include size of the bank, regulatory differences and the extent of moral hazard. After decomposing non-interest activities into three components (fee-based, trading and other activities), it is found that, for the Chinese banking market as a whole, fee-based and other income activities lead to a diversification discount, and that only G-SIBs can obtain benefits from diversification towards fee and commissions. However, trading activities can always improve banks’ performance level, for both the entire Chinese banking sector and its three sub-groups. Therefore we suggest that banks’ managers and regulatory authorities should set different diversification strategies and regulatory policies based on these diversification differences, so as to ensure that both specific banks and the entire banking system can maintain a higher profitability level.

References Allen F, Santomero AM (2001) What do financial intermediaries do? J Bank Financ 25(2):271–294 Acharya VV, Saunders A, Hasan I (2002) The effects of focus and diversification on bank risk and return: evidence from individual bank loan portfolios. Working paper Acharya V, Hasan I, Saunders A (2006) Should banks be diversified? evidence from individual bank loan portfolios. J Bus 79(3):1355–1412 Agoraki MEK, Delis MD, Pasiouras F (2011) Regulations, competition and bank risk-taking in transition countries. J Financ Stab 7(1):38–48 Allen L, Jagtiani J (2000) The risk effects of combining banking, securities, and insurance activities. J Econ Bus 52(6):485–497 Ansoff HI (1957) Strategies for diversification. Harv Bus Rev 35(5):113–124 Arellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev Econ Stud 58(2):277–297 Arellano M, Bover O (1995) Another look at the instrumental variable estimation of errorcomponents models. J Econom 68(1):29–51

86

3 The Impact of Income Diversification on Chinese Banks …

Back K, Brown DP (1993) Implied probabilities in GMM estimators. J Econom Soc 971–975 Baele L, De Jonghe O, Vander Vennet R (2007) Does the stock market value bank diversification? J Bank Financ 31(7):1999–2023 Barney JB (2002) Strategic management: from informed conversation to academic discipline. Acad Manag Perspect 16(2):53–57 Berger AN, Bouwman CH (2013) How does capital affect bank performance during financial crises? J Financ Econ 109(1):146–176 Berger AN, DeYoung R, Genay H, Udell GF (2000) Globalization of financial institutions: evidence from cross-border banking performance. Working paper Berger AN, Hasan I, Zhou M (2010) The effects of focus versus diversification on bank performance: evidence from Chinese banks. J Bank Financ 34(7):1417–1435 Berger PG, Ofek E (1995) Diversification’s effect on firm value. J Financ Econ 37(1):39–65 Blundell R, Bond S (1998) Initial conditions and moment restrictions in dynamic panel data models. J Econom 87(1):115–143 Boyd JH, Runkle DE (1993) Size and performance of banking firms: testing the predictions of theory. J Monet Econ 31(1):47–67 Boyd JH, Graham SL, Hewitt RS (1993) Bank holding company mergers with nonbank financial firms: effects on the risk of failure. J Bank Financ 17(1):43–63 Busch R, Kick T (2009) Income diversification in the German banking industry. Bundesbank Series 2 Discussion Paper No. 09/2009 Calmès C, Théoret R (2014) Bank systemic risk and macroeconomic shocks: Canadian and US evidence. J Bank Financ 40:388–402 Campa JM, Kedia S (2002) Explaining the diversification discount. J Financ 57(4):1731–1762 Canals J (1994) Competitive strategies in European banking. Oxford University Press, Oxford Casu B, Dontis CP, Staikouras S, Williams J (2016) Diversification, size and risk: the case of bank acquisitions of nonbank financial firms. Europ Financ Manag 22(2):235–275 Chi G-T (2006) Research on the relationship between income structure and income efficiency of China’s commercial banks. J Syst Eng 21:574–590 Chiorazzo V, Milani C, Salvini F (2008) Income diversification and bank performance: evidence from Italian banks. J Financ Serv Res 33(3):181–203 Claessens S, Demirgüç-Kunt A, Huizinga H (2001) How does foreign entry affect domestic banking markets? J Bank Financ 25(5):891–911 Comment R, Jarrell GA (1995) Corporate focus and stock returns. J Financ Econ 37(1):67–87 Cornett MM, Guo L, Khaksari S, Tehranian H (2010) The impact of state ownership on performance differences in privately-owned versus state-owned banks: an international comparison. J Financ Intermed 19(1):74–94 Danila N (2013) The new business model in the banking sector and its challenges. Strateg Think Chang World 79–96 De Haas R, Van Lelyveld I (2010) Internal capital markets and lending by multinational bank subsidiaries. J Financ Intermed 19(1):1–25 Deng X, Li S (2006) An empirical study of income diversification and performance: evidence from Chinese commercial bank. J Spec Zone Econ 9:339–341 (in Chinese) DeYoung R, Roland KP (2001) Product mix and earnings volatility at commercial banks: evidence from a degree of total leverage model. J Financ Intermed 10(1):54–84 Diamond DW (1984) Financial intermediation and delegated monitoring. Rev Econ Stud 51(3):393– 414 Drake L, Hall MJ, Simper R (2009) Bank modelling methodologies: a comparative non-parametric analysis of efficiency in the Japanese banking sector. J Int Financ Mark Inst Money 19(1):1–15 Easley D, O’hara M (2004) Information and the cost of capital. J Financ 59(4):1553–1583 Elsas R (2005) Empirical determinants of relationship lending. J Financ Intermed 14(1):32–57 Elsas R, Hackethal A, Holzhäuser M (2010) The anatomy of bank diversification. J Bank Financ 34(6):1274–1287

References

87

Esho N, Kofman P, Sharpe IG (2005) Diversification, fee income, and credit union risk. J Financ Serv Res 27(3):259–281 Fisher JG, Maines LA, Peffer SA, Sprinkle GB (2002) Using budgets for performance evaluation: effects of resource allocation and horizontal information asymmetry on budget proposals, budget slack, and performance. Account Rev 77(4):847–865 Flannery MJ (1989) Capital regulation and insured banks choice of individual loan default risks. J Monet Econ 24(2):235–258 Gamra SB, Plihon D (2011) Revenue diversification in emerging market banks: implications for financial performance. Unpublished manuscript Griliches Z, Hausman JA (1986) Errors in variables in panel data. J Econom 31(1):93–118 Gurbuz AO, Yanik S, Ayturk Y (2013) Income diversification and bank performance: evidence from Turkish banking sector. J BRSA Bank Financ Mark 7(1):9–29 Harvey CR, Zhou G (1993) International asset pricing with alternative distributional specifications. J Empir Financ 1(1):107–131 Holtz-Eakin D, Newey W, Rosen HS (1988) Estimating vector autoregressions with panel data. Econom: J Econom Soc 1371–1395 Hoskisson RE, Hitt MA (1990) Antecedents and performance outcomes of diversification: a review and critique of theoretical perspectives. J Manag 16(2):461–509 Jagtiani J, Nathan A, Sick G (1995) Scale economies and cost complementarities in commercial banks: on-and off-balance-sheet activities. J Bank Financ 19(7):1175–1189 Khanna T, Yafeh Y (2005) Business groups and risk sharing around the world. J Bus 78(1):301–340 Köhler M (2014) Does non-interest income make banks more risky? retail-versus investmentoriented banks. Rev Financ Econ 23(4):182–193 Köhler M (2015) Which banks are more risky? the impact of business models on bank stability. J Financ Stab 16:195–212 Kwan SH, Laderman ES (1999) On the portfolio effects of financial convergence-a review of the literature. Econ Rev 2:18–31 Landskroner Y, Ruthenberg D, Zaken D (2005) Diversification and performance in banking: the Israeli case. J Financ Serv Res 27(1):27–49 Lang HP, Stulz RM (1994) Tobin’s q, diversification and firm performance. J Polit Econ 102:1248– 1280 Lepetit L, Nys E, Rous P, Tarazi A (2008) Bank income structure and risk: an empirical analysis of European banks. J Bank Financ 32:1452–1467 Li L, Zhang Y (2013) Are there diversification benefits of increasing noninterest income in the Chinese banking industry? J Empir Financ 24:151–165 Lins KV, Servaes H (2002) Is corporate diversification beneficial in emerging markets? Financ Manag 31(2):5–31 Lou YC (2008) A study of commercial banking’s performance. Stud Int Financ 6:4–11 (in Chinese) Lown CS, Osler CL, Sufi A, Strahan PE (2000) The changing landscape of the financial services industry: what lies ahead?. FRB of N Y Econ Policy Rev (4) Massa M, Rehman Z (2008) Information flows within financial conglomerates: evidence from the banks–mutual funds relation. J Financ Econ 89(2):288–306 Mercieca S, Schaeck K, Wolfe S (2007) Small European banks: benefits from diversification? J Bank Financ 31(7):1975–1998 Mergaerts F, Vander VR (2016) Business models and bank performance: a long-term perspective. J Financ Stab 22:57–75 Miller DJ (2004) Firms’ technological resources and the performance effects of diversification: a longitudinal study. Strateg Manag J 25(11):1097–1119 Morgan DP, Samolyk K (2003) Geographic diversification in banking and its implications for bank portfolio choice and performance. In: Federal Reserve Bank of New York Working Paper Mulwa IM, Tarus D, Kosgei D (2015) Commercial bank diversification: a theoretical survey. Int J Res Manag Bus Stud 2(1):45–49

88

3 The Impact of Income Diversification on Chinese Banks …

Nguyen M, Perera S, Skully M (2016) Bank market power, ownership, regional presence and revenue diversification: evidence from Africa. Emerg Mark Rev 27:36–62 Nisar S, Peng K, Wang S, Ashraf B (2018) The impact of revenue diversification on bank profitability and stability: empirical evidence from south Asian countries. Int J Financ Stud 6(2):40–55 Palich LE, Cardinal LB, Miller CC (2000) Curvilinearity in the diversification–performance linkage: an examination of over three decades of research. Strateg Manag J 21(2):155–174 Petersen MA, Rajan RG (1995) The effect of credit market competition on lending relationships. Q J Econ 110(2):407–443 Ramakrishnan RT, Thakor AV (1984) Information reliability and a theory of financial intermediation. Rev Econ Stud 51(3):415–432 Sanya S, Wolfe S (2011) Can banks in emerging economies benefit from revenue diversification? J Financ Serv Res 40(1–2):79–101 Sargan JD (1958) The estimation of economic relationships using instrumental variables. Econom: J Econom Soc 393–415 Stein JC (2002) Information production and capital allocation: decentralized versus hierarchical firms. J Financ 57(5):1891–1921 Stiroh KJ (2004) Diversification in banking: is noninterest income the answer? J Money Credit Bank 36(5):853–882 Stiroh KJ, Rumble A (2006) The dark side of diversification: the case of US financial holding companies. J Bank Financ 30(8):2131–2161 Villalonga B (2004) Does diversification cause the “diversification discount”?. Financ Manag 5–27 Wagner W (2010) Diversification at financial institutions and systemic crises. J Financ Intermed 19(3):373–386 Zhou HW, Wang J (2008) A review of the fluctuation of non-interest income of commercial banks of our country from the perspective of portfolio theory. Econ Surv 4:044 (in Chinese)

Chapter 4

The Impact of Income Diversification on Chinese Banks: Bank Risk

Abstract This chapter investigates to what extent income diversification affects the risks of Chinese banks. First, this research measures both idiosyncratic risk and financial distress for specific banks. Then, it adopts the first-differenced GMM method based on the threshold dynamic panel model to investigate the income diversification effects on banks’ risk level.

In Chap. 3, an investigation focused on the diversification-performance nexus found that there exists an overall harmful diversification effect in the Chinese banking market. The Chapter also identified that, in addition to the in-depth consideration of profitability, banks engaging in income diversification must give due attention to the key aspect of safeguarding with regard to banks’ idiosyncratic risk and banks’ financial distress. In this chapter assesses the effects of income diversification on risk among Chinese banks. Using the threshold dynamic panel estimator based on the first-differenced GMM method, we find that for the overall Chinese banking sector, results reveal the existence of an inverse U-shaped relationship between bank risk and income diversification in terms of both banks’ idiosyncratic risk and banks’ financial distress.

4.1 Introduction Non-interest activities have become an important source of revenue for banks in mature as well as emerging markets in recent decades (Stiroh 2004; Laeven and Levine 2007; Sanya and Wolfe 2011, DeYoung and Torna 2013; Doumpos et al. 2016). Lately, bank diversification has also gained momentum in China. With the gradual development of market-oriented financial reform, Chinese banks are now actively engaged in non-interest business, including underwriting, brokerage and fiduciary services (Berger et al. 2010a; Li and Zhang 2013; Chen et al. 2017). As a result, while the traditional lending business continues to be their main source of revenue, Chinese banks are steadily shifting towards a multiple-revenue structure. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0_4

89

90

4 The Impact of Income Diversification on Chinese Banks …

Given the growing importance of business diversification for Chinese banks, it is imperative to study the risk implications of such a significant development. While the prior literature has indicated diversification can be beneficial for banks, debates remain as to whether and how the diversification-risk nexus are affected when banks move into non-traditional businesses. Earlier studies find a potential risk reduction from increasing the proportion of non-interest income (DeYoung and Roland 2001; Stiroh and Rumble 2006). However, recent research suggests that the inter-relatedness among banks’ non-interest products and services could mean a higher correlation, so a cross-selling strategy would result in excessive risk that could not be offset by portfolio diversification (Chen and Zeng 2014). Moreover, noninterest income may also mean other risk-enhancing factors, such as that they could be more volatile owing to high-financial leverage (DeYoung and Roland 2001), that there could be increased asymmetric information between the bank and borrowers (Mercieca et al. 2007), and that it is easy for customers of non-interest activities to switch to another bank (Li and Zhang 2013). This chapter aims to foster a better understanding of banks’ diversification and risk nexus through the case of Chinese banks. Existing studies focus mainly on mature markets. Few studies have focused on the diversification effect in the Chinese banking industry. This leaves a critical void in the literature, particularly in light of the growing importance of the nation’s banking industry, which has surpassed Europe to become the world’s second largest, following that of the US. We extend the current literature by developing an advanced methodology to address the particularity of China’s institutional background and data characteristics. Existing studies on bank diversity and risk have included very few considerations of dynamics in their relationship, with most adopting a static panel data model using either ordinary least squares regression (Berger et al. 2010b) or fixed effect estimation (Zhou 2014). Nevertheless, this research shows banks’ risk characteristics can generate dynamic changes. Specifically, we employ the first-differenced GMM estimator of dynamic panel models with threshold effects. This estimator addresses the inconsistencies generated from the endogeneity problem, providing a non-linear view of the dynamic GMM (Hsiao and Zhang 2015). The method addresses the endogeneity of both regressors and transition variables that are unlikely to be achieved by employing the standard least squares approach (Seo and Linton 2007). Moreover, this approach avoids the bias from quadratic terms that are widely used in the field; see Acharya et al. (2006), Gamra and Plihon (2011) and Brei and Yang (2015). Our findings show the existence of a non-linear diversification effect on banks’ idiosyncratic risks and the probability of banks’ financial distress and to several reasons. Consistent with Acharya et al. (2006), a bank’s monitoring effectiveness might be lower in newly entered and competitive sectors, so the diversification may initially lead to a poorer quality of the loan portfolio and higher potential operational risk. Over time, however, this diversification discount will gradually change to a diversification benefit. Second, the inverted U-shaped relationship may also be caused by the learn-by-doing effect. Learning can be a dynamic process, and banking institutions gradually understand specific characteristics of the business lines and the risk levels thereof, and thus, what proportion of non-interest income is most suitable

4.1 Introduction

91

to them (Gamra and Plihon 2011). With richer experience, the benefits of diversification across various sources of earnings would offset the costs of increased complexity and the associated idiosyncratic risk. Size does matter for banks diversifying into new businesses other than traditional interest earning activities (Chiorazzo et al. 2008; De Jonghe et al. 2015). Our sample comprises 35 Chinese commercial banks and are divided into three sub-groups: Global Systemically Important Banks (G-SIBs), Domestic Systemically Important Banks (D-SIBs), and banks that are not classified by the authorities as systemically important (N-SIBs) to capture this effect. The remainder of the chapter is organized as follows. Section 4.2 provides an overview of relevant literature on bank risk and income diversification. The data on variables and the methodology are presented in Sect. 4.3. Section 4.4 outlines the model specification and reports the empirical results. Section 4.5 offers concluding remarks.

4.2 Related Literature Theoretically, diversification of income sources can improve the stability of banks’ cash flows and disperse their idiosyncratic risks. Stiroh (2004) maintains that the traditional lending business provides a channel for banks to attract clients to their non-interest activities, as people are more likely to seek fee-based services in the same bank. Wagner (2010) shows that banks are keen to adopt a strategy that uses attractive lending and deposit rates to improve customer stickiness and to make themselves more profitable through high-return non-interest income. Pennathur et al. (2012) suggest that combining the two business assets into a portfolio can be mutually beneficial and there would not be crowd-out effects between them. Given that noninterest and interest incomes are not perfectly correlated, a portfolio of these activities can be risk reducing. In addition to the portfolio effect, diversified banks can also obtain rich information from their mixed business lines. Use of the information can help banks improve risk management (Stein 2002; Elsas et al. 2010). However, some authors argue that fee-earning activities could be associated with higher risk than are interest activities, as they will increase the overall volatility of the portfolio (Stiroh and Rumble 2006; Demirgüç-Kunt and Huizinga 2010; Köhler 2014). One reason for the increase in volatility is that income from non-interest activities may have greater fluctuation, because it is easier for clients to switch between banks in these activities than in lending activities (Lepetit et al. 2008). In the face of cyclical fluctuation, non-interest income would decrease dramatically, while banks with a larger proportion of interest activities might maintain a more stable performance compared with more diversified banks (Köhler 2014). Another reason is that non-interest activities require less regulatory capital, which may lead banks to have a higher degree of financial leverage and, hence, may increase earnings volatility (DeYoung and Roland 2001). In addition, an agency problem may also play a role

92

4 The Impact of Income Diversification on Chinese Banks …

here. Given interest conflicts with shareholders, managers have incentives to overdiversify, which will, in turn, harm the wider financial system because diversification makes institutions become more similar to each other by exposing them to the same risks, which may cause a joint failure (Acharya and Yorulmazer 2008; Wagner 2010). Empirical evidence on whether fee-earning activities would increase banks’ risk has been mixed. Chiorazzo et al. (2008) show positive evidence suggesting that diversification can improve the trade-off between risk and income for Italian banks. Köhler (2014, 2015) finds that banks’ earnings will be more stable and profitable if they diversify into noninterest income. Similar results are also reported in Mergaerts and Vander (2016) and Nguyen et al. (2016). Recent research suggests that revenue diversification could also be beneficial to banks from developing countries (Sanya and Wolfe 2011). Conversely, negative evidence has been found by an extensive body of research. Many studies focusing on US banks find that diversification fails to produce a greater performance. Such research includes DeYoung and Roland (2001), DeYoung and Rice (2004), Stiroh (2004), Stiroh and Rumble, (2006), Acharya et al. (2006), Hayden et al. (2007), Mercieca et al. (2007), Goddard et al. (2008), and Berger et al. (2010a, b). For risk implications, evidence of an adverse relationship between diversification and risk of American banks has been presented in, for example, Demsetz and Strahan (1997), and DeYoung and Roland (2001). Baele et al. (2007), analysing a panel data of banks during 1989–2004, find that a higher share of non-interest income positively affects bank’ franchise values, but increases their systematic risk. Lepetit et al. (2008) suggest that diversified banks present higher risk than banks mainly conducting traditional business, though this positive link between diversification and risk is largely a phenomenon of small banks and driven by commission activities. Another strand in the empirical literature explores the phenomenon whereby, although the non-interest income as a whole may lead to benefits or discount, the bank risk estimation varies among different activities under non-interest income. Stiroh (2004) divides non-interest income into three components, namely, fee-for-service income, trading revenue, and other types of non-interest income, and suggests that these different activities could have different effects on risk. Lee et al. (2014) consider 29 Asia-Pacific banking markets and identify that risk derives mainly from commission and fee activities, and that trading and other non-interest incomes could reduce banks’ risk level. Similarly, Meslier et al. (2014) investigate an emerging country sample from 1999 to 2005. They estimate valuation variability of a bank’s stock returns and find that increasing the share of fee-based income leads to a risk increase. From the view of Elyasiani and Wang (2008), such fee-for-service income makes banks less transparent to investors and, thus, makes the task of bank supervision more difficult. Recently, researchers have started to examine whether the effect of diversification differs according to bank-specific characteristics, such as size and types. By using the GMM system, Goddard et al. (2008) suggest that a similar cross-selling strategy is not suitable for both large and small credit unions. With the same method, Chen and Zeng (2014) show that small-sized banks in the EU banking industry are restricted to selected market segments; thus, smaller banks may be associated with a higher

4.2 Related Literature

93

risk exposure and higher default risk than are larger banks. Similarly, Köhler (2014) find that larger banks are more likely to be active in volatile and risky trading and off-balance sheet activities, which allows them to employ a higher financial leverage than small banks. In addition, some studies also claim that size increase due to diversification could make the financial system more fragile. Small-sized banks are highly susceptible to the failure of large banks, which poses a high external risk for the entire financial sector (Kobayashi 2012). However, as suggested by Gurbuz et al. (2013), large-sized banks generally have better information technology, human capital management and risk management. Therefore, diversified bank with larger size could cause technology spillover and create advantages from the economy of scope, thus outperforming others within the banking industry (Mercieca et al. 2007). Mixed findings on the risk profiles of banks’ revenue diversification suggest that while diversification into fee-earning activities may be beneficial to some banks, it has a dark side since the volatile non-interest income may offset the benefits due to the portfolio effect (Stiroh and Rumble 2006). In this light, further empirical investigation into the effect of revenue diversification on bank risk would be desirable and necessary. The static panel data analysis is based on both accounting and stock market data, and the estimation methods used ranges from pooled OLS (Lepetit et al. 2008; Kobayashi 2012) to fixed effects (Bonin et al. 2005; Tabak et al. 2011) and random effects GLS estimation (Mergaerts and Vander 2016). However, the approach suffers from the endogeneity problem because diversification strategies are correlated with banks’ business opportunities (Acharya et al. 2006; Stiroh and Rumble 2006; Baele et al. 2007; Busch and Kick 2009, Sanya and Wolfe 2011; Köhler 2014). In 1995, Berger and Ofek identified a diversification discount without controlling for endogeneity. Subsequent papers, however, find that with consideration of the endogeneity problem, they obtain an inverse result; that is, the result becomes positive with the same methodology (Campa and Kedia 2002; Villalonga 2004). To address endogeneity and other estimation inconsistences in the banking context, many studies then select to employ dynamic panel models, or the generalized methods of moments (GMM) estimator. Research in this strand of the literature includes Sanya and Wolfe (2011), Nguyen (2012), Vallascas and Keasey (2012) and Gambacorta et al. (2014) to name just a few. Recent studies noted the existence of a non-linear relationship between diversification and banks’ risk (Demirgüç-Kunt and Huizinga 2010; Gamra and Plihon 2011). A common method to capture such a non-linear relationship is to include a quadratic term in the empirical model (Gambacorta et al. 2014; Brei and Yang 2015). Acharya et al. (2006) add a risk-squared variable (Risk 2 ) in the regression specification. Baele et al. (2007) introduce the square of non-interest revenue share into their empirical model. Importantly, these models shed light on the nonlinear nature of the banks’ diversification-risk nexus, but the banks’ income diversification as a process could be not only nonlinear but also dynamic. However, these terms are exogenously given in nature, and may thus be inadequate for treating the endogeneity problem. Furthermore, it is plausible that diversification would only start to exert its effects beyond

94

4 The Impact of Income Diversification on Chinese Banks …

some threshold levels. This finding implies that consideration of the threshold effect in a dynamic nonlinear model is also required.

4.3 Data and Methodology 4.3.1 Data Sample Our chapter sample comprises an unbalanced panel of annual report data from 35 Chinese banks over the period from 2007 to 2016. Following the Guidelines for the Disclosure of Global Systemic Importance Indicators of Commercial Banks issued by the China Banking Regulatory Commission (CBRC), these commercial banks can be divided into three groups: global systemically important banks (G-SIBs), domestic systemically important banks (D-SIBs), and other banks (N-SIBs). The reason for choosing 35 banks from 2007 to 2016 rather than 40 banks from 2005 to 2015 as in Chaps. 3 and 5 is that the first-differenced GMM estimator for the dynamic threshold panel data model requires balanced data. Of the 40 banks used in those chapters, five did not exist before 2007. Therefore, in order to maximize the data, the observation period is reduced by two years, and the sample includes only those banks that were established in or before 2007. In this chapter, all balance sheet and income information used to construct the variables for empirical analysis are taken from Bankscope. Daily market and bank stock returns are taken from DataStream International. Industry-specific and macroeconomic data are obtained from the National Bureau of Statistics of China.

4.3.2 Variables (1) Dependent Variables: Idiosyncratic Risk Measures First this chapter adopts two accounting-based idiosyncratic risk indicators which are widely used in the empirical banking literature, namely, the ratio of loan loss provisions and customer loans (LLP) and the ratio of impaired loans and equity (ILE). As suggested by Hsieh et al. (2013), with such a traditional measure, a higher value indicates that the bank has higher level of risk for its loan portfolio. In order to find out which specific risk would be affected by income diversification, we also divided the non-systemic risk into three catalogues including credit, liquidity and interest rate risks. Following Valverde and Fernández (2007), we also divided the non-systemic risk into three categories, including credit, liquidity and interest rate risks. Credit risk (CREDIT) is calculated as the one-lagged ratio of loan default to total loans. Liquidity risk (LIQUIDITY) is the ratio of liquidity assets to short term

4.3 Data and Methodology

95

funding, and interest rate risk (INTEREST) refers to the changes in the three-month interbank market rate. (2) Dependent Variables: Financial Distress Measures Recent studies of diversification effect also tend to take account of financial distress (such as DeYoung and Torna 2013), in order to achieve a deep understanding of banks’ default and insolvency risk. Two commonly used measures to detect such financial distress are the Z-score and distance to default. The former is a popular measure of soundness of a bank because it combines the bank’s buffers such as capitalization with returns and volatility, and hence allows investigation of bank risk and stability. Studies employing the Z-score to measure the probability of bank insolvency and bankruptcy include Berger et al. (2009), Laeven and Levine (2009), Angkinand and Wihlborg (2010), Barry et al. (2011), and Demirgüç-Kunt and Huizinga (2010). Following Lepetit et al. (2008), the original Z-score can be extended to become the ZP-score by including a risk index to investigate various bankruptcy-related scenarios. The DD indicator quantifies bank distress by gauging how far a firm is from default and is widely used for measuring default risk (Vassalou and Xing 2004; Hovakimian et al. 2012). Recently, it has also been applied to check the soundness of Chinese banks (Chen et al. 2010). As a risk predictor, it has been verified as useful by numerous studies and its predictive power with regard to bank fragility outperforms that of several accounting and market-based measures (Campbell et al. 2008; Bharath and Shumway 2008). The DD index is generally calculated according to the KMV-Merton model for listed banks and Private Firm Model (PFM) separately. Both distance to default and ZP-score have an inverse correlation with financial distress and banks with a higher DD and ZP-score are less likely to default. Therefore, our study uses the opposite of DD and ZP-score to measure the risk level which would have the same direction with that of LLP and other risk indicators. (3) Independent Variables: Measures of Income Diversification Similarly to Mercieca et al. (2007), our study measures the level of income diversification by constructing a Herfindahl-Hirschman index (HHI). We also divide the non-interest income into three categories: revenue from trading in foreign exchange and fiduciary activities; fee and commission income gained from clearing, settlement and other financial services; and other non-interest income. Our construction of the HHI index follows Elsas et al. (2010). Conversely, the diversification levels are measured respectively along the three categories of income. Following Köhler (2014), we also construct the diversification index by employing the ratio of net fee and commission income to total operating income (COM), the ratio of net trading income to total operating income (TRA) and the ratio of net other operating income to total operating income (OTH).

96

4 The Impact of Income Diversification on Chinese Banks …

(4) Control Variables LCD: This variable is the ratio of loans to customer deposits. According to Cornett et al. (2010), this ratio can reflect the level of banks’ liquidity. The larger the ratio is, the lower the level of liquidity will be. ETA: To adjust for banks’ attitude towards risk, we adopt the ratio of equity over assets, which describes the degree of total financial leverage and capital adequacy (Stiroh 2004; Pennathur et al. 2012; Gurbuz et al. 2013). According to Busch and Kick (2009), a well-capitalized bank is less likely to become insolvent and more likely to be engaged in low-risk investment to ensure that it operates soundly. CIR: Cost-income ratio is estimated through the operating expenses relative to gross income which measuring banks’ cost structure (Busch and Kick 2009).

4.3.3 Methodology Prior research in the field is mostly focused in static analysis. However, banks’ income diversification is a long process, during which the evolvement of the diversificationrisk nexus could be dynamic and non-linear. In addition, there may exist a threshold effect since it is plausible that only beyond some threshold level would diversification start to have a significant effect on bank risk. This finding implies that the static and linear method could be inadequate for capturing the risk implications of diversification. Rather, a dynamic threshold model would be more fitting. In this study, a GMM-type threshold model, or the first-differenced GMM estimator for the dynamic threshold panel data model, as proposed by Seo and Shin (2016), is employed. This approach has the advantage of giving a dynamic view of the diversification effects on bank risk while avoiding both the endogeneity problem usually associated with the static model and the bias from using the quadratic terms as proxy for the non-linear relationship. It can also overcome the problem associated with previous GMM-type threshold models, whereby both regressions and threshold variables must be exogenous (Ramírez-Rondán 2015). The research proceeds by assuming nonlinearity in Chinese banks’ diversificationrisk nexus for two reasons. First, Chinese banks on the whole are still in the early stages of diversification. When a bank first begins to engage in non-interest activities, managers generally lack the necessary skills and professionalism. Given the circumstances, some decisions may be irrational, exposing the bank to excessive risks (Deng and Li 2006). Second, with under-developed governance structure, moral hazard is more likely to emerge, leading to inefficient allocation of resources and overdiversification (Barry et al. 2011). However, over time, this excess may be gradually mitigated with both the expanding of diversification and obtaining of better information among a bank’s management. Consequently, the diversification effects can be non-linear and vary with the changing proportions of non-interest activities.

4.3 Data and Methodology

97

Conventional econometric approaches to capturing such a non-linear relationship are to catalogue different groups with 25th, 50th, and 75th percentiles of non-interest income shares (Chiorazzo et al. 2008) or to include quadratic terms in empirical models (e.g., Gambacorta et al. 2014; Brei and Yang 2015). However, use of such proxies cannot provide consistent results in the face of a certain threshold value. Rather, a threshold model would be more appropriate in that it treats the sample split value as unknown. Tong (1978) first developed the threshold auto-regression (TAR) model for use in time-series analysis. This method estimates the threshold variables to determine the threshold point and to avoid the bias generated from using a subjective approach to determine the critical value for each group. The static threshold model has subsequently been developed to cover cross-sectional and panel data (e.g., Tiao and Tsay 1994; Martens et al. 1998; Hansen 1999). The most widely used threshold estimation, developed by Hansen (1999), employs a framework with a panel dataset {yit , qit , xit : 1 ≤ i ≤ n, 1 ≤ t ≤ T }, where yit refers to the dependent variable, qit is a scalar of threshold variable and xit represents the vector of all control variables included in the regression. In setting this model, all regressors, including the threshold variables, are required to be exogenous. To estimate the regression slope, Hansen (1999) adopts the OLS estimation with restrictions on xit and qit being time variant. Research finds high-risk banks are more likely to enter into riskier industries and produce riskier loans (Acharya et al. 2006). This endogeneity bias proves to exist both in the decision to diversify and in systematic differences between different types of banks. Based on Hansen (2000), Caner and Hansen (2004) develop an asymptotic framework and employ a two-stage least square estimation to address the endogeneity problem. Subsequent studies have followed this two-stage least square method. However, the improvements offered by this method remain limited in that both regressors and threshold variables have to be exogenous (Seo and Linton 2007; Yu 2012). In response, recently several studies have attempted to address this endogeneity problem of threshold variables by employing the dynamic threshold regression model. Kourtellos et al. (2016) constructed a two-stage concentrated least squares method. Ramírez-Rondán (2015) proposed a maximum likelihood estimation of the threshold with slope parameters. In our study, the non-linear effect is estimated by using the first-differenced GMM threshold dynamic panel model based on Seo and Shin (2016). This method provides a dynamic and non-linear view over the relationship between income diversification and risk. By adopting the two-step first-differenced GMM estimator, it overcomes the endogeneity bias and allows for both regressors and threshold variables to be endogenous. In order to estimate the persistence of the diversification level, it is necessary to employ a threshold model that accounts for endogeneity and dynamic effect. Seo and Shin’s (2016) approach applies the GMM estimation techniques to a dynamic panel estimation, which is based on the static threshold model proposed by Hansen (1999) and a modified 2SLS-like threshold panel regression method introduced by Caner and Hansen (2004).

98

4 The Impact of Income Diversification on Chinese Banks …

The starting point of dynamic threshold analysis is the specification of a static threshold model:       yit = 1, xit φ1 1(qit ≤ γ ) + 1, xit φ2 1(qit > γ ) + εit , i = 1, . . . , n; t = 1, . . . , T

(4.1)

where yit refers to a scalar stochastic variable, which in our study can be defined as income diversification; xt is the k1 × 1 vector of time-varying regressors allowing for endogeneity and lagged dependent variable; 1 {·} is an indicator function; qit is the transition variable, and γ is the threshold parameter; φ1 and φ1 are slope parameter falls in the upper and lower regimes respectively; εit refers to the regression error, which consists of two components: εit = αi + vit ,

(4.2)

where αi is an unobserved individual fixed effect and vit is a zero mean idiosyncratic random disturbance and is assumed to be a martingale difference sequence, E(vit |Ft−1 ) = 0

(4.3)

where Ft−1 refers to a natural filtration at time t. In line with the Seo and Shin (2016) approach, there is no assumption that threshold variable (qit ) and regressors (xit ) are measureable with respect to Ft−1 . Therefore, both qit and xit can be endogenous. In addition, both “fixed threshold effect” and “diminishing or small threshold effect” are allowed for statistical inference of the threshold parameter (γ ) by defining: δ = δn = δ0 n −α for 0 ≤ α < 1/2.

(4.4)

In order to solve the correlation of the regressors with individual effects in Eqs. 4.1 and 4.2, Seo and Shin (2016) utilise the GMM estimator introduced by Arellano and Bond (1991). Equation 4.1 can be rewritten by considering the first-difference transformation as: 

yit = β  xit + δ  X it 1it (γ ) + εit ,   where β = φ12 , . . . , φ1,k+1 , k1 ×1

δ

(k1 +1)×1

(4.5)

= φ2 − φ1 , and

       1it (γ ) 1(qit > γ ) 1, xit   = X it = and .  −1(qit−1 > γ ) 2×1 1, xit−1 2×(1+k1 )

(4.6)

  where θ = β  , δ  , γ , and θ belongs to a compact set, Θ = Φ × Γ ⊂ Rk , combined with k = 2k1 + 2.

4.3 Data and Methodology

99

As GMM estimator is used in this dynamic threshold approach, it is necessary to define instrument variables. Where, the 1 × 1 vector of instrument variables,      z it0 , . . . , z i T for 2 < t0 ≤ T , such that either     E z it0 εit0 , . . . , z i T εi T = 0,

(4.7)

or for each t = t0 , . . . , T, E(εit |z it ) = 0,

(4.8)

In order to estimate parameters, Seo and Shin (2016) apply a two-step GMM estimation procedure for the estimation of non-linear relation. First, for the value of a selected threshold variable γ, θ = (φ, β1 , β2 , δ) is estimated through the firstdifference GMM by using the instruments suggested by Arellano and Bond (1991). Second, this estimation is repeated for γ’s belonging in a strict subset of the support of the threshold variable. Therefore, it is possible to calculate a different θˆ for each selected γ. Under this condition, the parameter value of γ could minimize the GMMtype objective function; thus θˆ can be defined as the two-step optimal GMM estimator. In detail, 1 × 1 vector of the sample moment conditions can be written as: n 1 gi (θ ), g¯ n (θ ) = n i=1

(4.9)

where  ⎞  z it0 yit0 − β  xit0 − δ  X it0 1it0 (γ ) ⎜ ⎟ .. gi (θ ) = ⎝ ⎠. .      z i T yi T − β xi T − δ X i T 1i T (γ ) ⎛

(4.10)

      In Eq. 4.10, gi = gi (θ0 ) = z it0 εit0 , . . . , z i T εi T and Ω = E gi gi . Where, p

under the assumption in which Ω is finite and positive definite, Wn → Ω −1 . Given  J¯n (θ ) = g¯ n (θ ) Wn g¯ n (θ ).

(4.11)

The estimation of parameter θ can be estimated by minimizing J¯n (θ ): 

θ = arg min J¯n (θ ) θ∈Θ

(4.12)

Above GMM estimation overcomes the drawbacks of previous methods that fail to take account the endogeneity problem leading to biased estimation. Berger and Ofek

100

4 The Impact of Income Diversification on Chinese Banks …

(1995) identify a negative relation between the franchise value and income diversification level, without controlling for endogeneity. However, subsequent papers, such as Campa and Kedia (2002) and Villalonga (2004), find that with consideration of the endogeneity problem, the result becomes positive with the same methodology. In addition, some prior studies, such as Stiroh and Rumble (2006) and Baele et al. (2007), also maintain that control of endogeneity of diversification is required, because the diversified strategies are highly correlated with their business opportunities. Therefore, the GMM-threshold model is suitable and necessary for our study because of the possible existence of the endogeneity in the relation between income diversification and risk since more risky banks are more likely to diversify (Acharya et al. 2006). As such, GMM-threshold estimation is employed in this paper in a setting of the threshold model which allows for endogenous regressors and endogenous threshold variables.

4.4 Empirical Estimation and Results 4.4.1 Estimation Steps To estimate the parameters, Seo and Shin (2016) applied a two-step GMM estimation procedure for the estimation of a non-linear relationship. First, for the value of a selected threshold variable γ, θ = (φ, β1 , β2 , δ) is estimated through the firstdifference GMM by using the instruments suggested by Arellano and Bond (1991). Second, this estimation is repeated for γ’s belonging in a strict subset of the support of the threshold variable. Therefore, it is possible to calculate a different θˆ for each selected γ. Under this condition, the parameter value of γ could minimize the GMMtype objective function; thus, θˆ can be defined as the two-step optimal GMM estimator. The GMM-threshold estimator is employed in a non-linear setting, which allows for endogenous regressors and endogenous threshold variables. The estimator overcomes the drawbacks of previous methods that fail to take into account the endogeneity problem leading to biased estimation. Two tests are necessary for the model. First, a Hausman type test is used to check the validity of the over-identifying moment conditions. Under the null hypothesis, models have effective instruments and avoid over-identifying restrictions. In that case, instruments in the GMM models are appropriately orthogonal to the error and are therefore reasonable and statistically acceptable. Second, we test for linearity or threshold effects as a result of the presence of unidentified parameters.

4.4 Empirical Estimation and Results

101

4.4.2 Overall Results 4.4.2.1

Income Diversification and Risk: Whole Sample

Table 4.1 reports the regression results between the diversification level and four different bank risk measures. Coefficients of control variables appear largely reasonable. According to the regression results for variables with idiosyncrasy risk (Model 1 and Model 2), the threshold estimations are 20.166 and 12.158, respectively. One can see that the diversification level is positively correlated with banks’ idiosyncratic risk, while at the same time, higher levels of diversification are associated with a reduction in bank risk. In general, our study echoes the work of Gamra and Plihon (2011), who use data from both East Asian and Latin American markets and find that income diversification makes banks less stable in the early stages of engaging in mixed business lines, but that they become more stable with the expansion of non-interest activity. These results can be explained from several perspectives. First, the process of diversification generates moral hazard problems and monitoring difficulties. Especially in the early stages of diversification ineffective monitoring make it difficult for both insiders and outsiders to observe the running of bank operations, which will have adverse impacts on bank risk. As the diversification in non-interest activity develops and hence the proportion of non-interest over total operational income increases, banks become more motivated to continually improve the governance and supervision regime to ensure that non-interest activity is not used in ways that would deteriorate both specific risk and financial stability (Ashraf et al. 2016). Second, expansion of non-interest activities requires employees to have special knowledge, and the application of relatively advanced technology. Banks can reduce the risks through accumulation of sufficient experience and by studying the risk characteristics of non-interest activities and get benefits from leveraging managerial skills (Iskandar-Datta and McLaughlin 2007). Furthermore, several advanced noninterest activities, such as securitization, can reduce possible shortfalls on payments to debtholders. However, such activities can be launched and supply a buffer to absorb losses only when banks have enough professional employees. To date, very few such non-interest activities have been launched in the Chinese banking sector. On the other hand, it is likely that, with the accumulation of experience and increased diversification, such advanced non-interest activities could make banks safer and more stable. In this chapter, we also take account of the diversification effect on banks’ financial distress. Table 4.1 reports the results by using Model 3 and Model 4 with consideration of financial distress. We find that non-traditional banking activities have economically meaningful effects on the probability of bank failure, and again, the results confirm the inverse U-shaped relationship between diversification and bank risk. These results can be explained as follows. The calculation method for distance to default considers banks’ debt level and whether or not the cash flow state is healthy. In the early stages of expansion of non-interest activities banks incur high initial costs,

102

4 The Impact of Income Diversification on Chinese Banks …

Table 4.1 Income diversification and risk for Chinese banks from 2007 to 2016 Idiosyncratic risk

Financial distress

ILP

LLE

Z-score

DD

(1)

(2)

(3)

(4)

It-1

−0.457*

−1.142*

1.744***

5.049***

(0.252)

(0.630)

(0.493)

(1.931)

HHI

0.036

0.736*

0.067

1.072***

(0.025)

(0.379)

(0.049)

(0.399)

−0.116***

0.275

−0.010

−1.462***

(0.033)

(0.294)

(0.028)

(0.444)

ETA

0.090***

0.667*

0.092

0.174

(0.028)

(0.359)

(0.099)

(0.199)

CIR

−0.202**

1.999

−0.486

−1.344

(0.098)

(1.331)

(0.439)

(1.598)

1.351***

−3.607***

1.317***

−1.727***

(0.400)

(0.936)

(0.317)

(0.405)

−0.123***

−1.032**

−0.054**

−0.101

(0.037)

(0.480)

(0.022)

(0.237)

LCD

−0.030

1.547***

0.111***

−0.762**

(0.025)

(0.380)

(0.041)

(0.338)

ETA

0.047***

0.570

−0.026***

−0.853***

(0.018)

(0.391)

(0.099)

(0.181)

−0.468

1.645

−0.391

1.563

(0.333)

(1.684)

(0.365)

(1.456)

Threshold

20.166***

23.104***

19.048***

21.569***

(0.792)

(1.580)

(7.082)

(1.244)

J-statistic

0.443

0.616

0.165

0.356

Upper regime

0.273

0.174

0.329

0.223

Linearity test

0.014

0.001

0.022

0.000

Observations

315

315

315

315

Lower regime

LCD

Upper regime It-1 HHI

CIR

This table reports the first-differenced GMM threshold dynamic panel model developed by Seo and Shin (2016). Our dependent variables are loan loss provisions/customer loans (LLP), impaired loans/equity (ILE), opposite distance to default (DD) and opposite Z-score. It-1 refers to the lagged one period of the dependent variables. HHI indicates the income diversification using HerfindahlHirschman Index with the components of interest income and three activities under non-interest income. LCD is Loans/Customer Deposits, ETA refers to Equity/Total Assets, CIR is Cost to Income Ratio. J-statistic checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. Linearity test checks whether threshold exists. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level

4.4 Empirical Estimation and Results

103

for example for the establishment of infrastructure and electronic platforms, and for specialist staff training. However, according to the empirical results reported from Chap. 3, the non-interest income cannot make the overall Chinese banking industry more profitable. Therefore, the results of Models 3 and 4 reflect a situation in which, in the early stage, the process of income diversification will not bring benefits to alleviate banks’ financial distress. However, with an increased level of non-interest activities, the initial costs can bring sustained profit and cash flow. Meanwhile, banks’ debt is mainly generated from deposit and other interest businesses. Hence, the mixed business lines strategy also plays a role in reducing bank debt and increasing the overall liquidity level. Therefore, once the income diversification has passed a certain threshold, it makes banks more stable and reduces their financial distress.

4.4.2.2

Income Diversification and Risk: Evidence on China’s Global, Domestic and Non-systemically Important Banks

Table 4.2 reports the results from estimating the association between income diversification on one hand, and insolvency risk and financial stress on the other for G-SIBs, D-SIBs and N-SIBs during the period 2007–2016.1 The results from the GMM type dynamic threshold estimation indicate that diversification is an important determinant of banks’ risk indicators, but presents different diversification effects in different groups.

China’s Global Systemically Important Banks According to the results reported in Table 4.2, in contrast to the results for the whole sample, for G-SIBs the coefficients of the HHI are significantly and negatively associated with both idiosyncratic risks and financial distress. This is the case for both the lower and higher regimes. However, in the upper regime the coefficients of both risk indicators are higher than those in the lower regime. This indicates that Chinese G-SIBs will always gain risk-reduction benefits when diversifying into non-traditional businesses, but that after passing a certain threshold point of income diversification, risks will reduce even further, helping G-SIBs to improve their financial situation. Such diversification benefits are in line with our first empirical results, where the mean of HHI for G-SIBS is 25.125, which is higher than the threshold point for all four models in Table 4.2. Therefore we suggest that, unlike the other two groups, G-SIBs have already passed the threshold point from diversification discount to diversification benefits. However, greater diversification could lead to yet further reduction of risk. 1 As the small sample size might result in an incidental parameters problem, this book also applies a

robustness test by using dummy variables to the catalogue of the three sub-groups. The robustness test results are reported in the Appendix.

LCD

HHI

It-1

Upper regime

CIR

ETA

LCD

HHI

It-1

Lower regime (0.383)

−0.716* (0.933)

7.587***

(0.014) (0.005)

(0.053)

(0.061)

(0.259)

(0.014)

(0.255) (0.613)

−1.578**

(0.209)

0.746***

(15.556)

(0.034)

(0.304)

−0.241*** 0.704**

(0.091)

(0.025)

−0.052**

(0.005)

0.006

(0.009)

0.018**

(0.236)

1.101***

(3)

Z-score

(0.052)

−0.064

(0.05)

0.532***

(0.043)

0.047

(0.274)

0.187

(4)

DD

Financial distress

N-SIBs

(2)

LLE

(0.328)

0.501

(0.564)

(0.034)

0.162***

(0.057)

(0.273)

0.354

(0.212)

−0.182*** 0.431**

(0.049)

0.330***

(0.275)

−1.253*** −1.057*

(1)

ILP

Idiosyncratic risk

(0.546)

4.050***

(0.239)

(3.045)

5.742*

(0.173)

(1.924)

−3.925**

(1.31)

(0.02)

0.057***

(0.015)

0.027*

(0.168)

0.107

(4)

DD

(0.303)

−0.140

(0.150)

0.260*

(0.030)

(0.394)

−2.043***

(0.057)

-0.287***

(0.017)

−0.205*** 0.024

(0.062)

−0.124**

(0.126)

0.270**

(0.291)

0.059

(3)

Z-score

Financial distress

(0.004)

−0.001

(0.011)

(0.037)

0.335***

(0.096)

(0.082)

−0.125

(0.477)

(0.788)

1.603**

(0.568)

(0.015) (0.032)

(continued)

(0.009)

−0.138*** 0.028***

(0.052)

−0.063*** −0.253*** −1.304*** −2.066*** −0.165*** −0.030**

(0.248)

1.108***

(0.079)

−31.442** −0.540*** −3.806*** −0.698*** 1.973

(0.221)

(0.682)

3.035***

(0.150)

1.547***

(0.004)

(0.003)

(0.312)

(1.106)

−0.347

0.104***

(0.742)

(0.046)

−0.157*** −2.192*** −0.143

−0.057*** 2.521***

(0.005)

−0.063*** −0.623

(1.184)

−0.194*** 1.624

(1.990)

(0.061)

(0.057)

−0.335*** 6.945***

(0.082)

(0.047)

(0.007)

(0.012)

(0.011)

−11.08*** −0.481*** 8.078***

(0.001)

(0.015)

0.371***

1.717***

0.005***

(0.066)

−2.297*** −0.265*** 2.270**

(0.011)

−0.035*** −0.793*** −0.232*** −0.503*

−0.027**

(0.007)

(0.003)

(0.004)

−0.059*** 2.297***

(0.013)

(0.318)

(0.004)

−0.007**

(2)

(0.002)

(0.058)

0.692***

(1)

1.083***

(0.004)

(0.019)

(4)

−0.057*** −0.426*** −0.067*** −1.781*** 0.311***

1.616***

0.195***

(3)

(2)

(1)

LLE

LLE

ILP

ILP

Z-score

Idiosyncratic risk DD

D-SIBs Idiosyncratic risk

Financial distress

G-SIBs

Table 4.2 Income diversification and risk for Chinese G-SIBs, D-SIBs and N-SIBs, 2007–2016

104 4 The Impact of Income Diversification on Chinese Banks …

0.072

0.054

36

Linearity test

Observations

(0.053)

36

0.034

0.447

0.917

(4.021)

36

0.000

0.168

0.450

(0.059)

29.205***

(0.022)

81

0.000

0.277

0.978

(0.475)

22.901***

(0.300)

1.262***

(0.053)

0.271***

(4)

DD

N-SIBs

(0.154)

0.562***

(1)

ILP

(0.369)

81

0.056

0.251

0.493

(1.255)

22.458***

(0.177)

81

0.000

0.269

0.350

(2.564)

22.951***

(0.352)

81

0.005

0.276

0.872

(0.728)

22.902***

(1.652)

198

0.042

0.041

0.496

(0.236)

24.691***

198

0.000

0.098

0.381

(1.155)

26.870***

(2.412)

6.605***

(0.18)

0.451**

(2)

LLE

Idiosyncratic risk

−5.661*** −0.547*** −2.437*** 3.198*

(0.013)

0.028**

(3)

Z-score

Financial distress

198

0.000

0.69

0.578

(0.318)

9.218***

(0.168)

−0.169

(0.045)

0.027

(3)

Z-score

198

0.051

0.333

0.262

(0.653)

15.183***

(0.071)

0.162**

(0.009)

0.048***

(4)

DD

Financial distress

This table reports the first-differenced GMM threshold dynamic panel model developed by Seo and Shin (2016). Our dependent variables are loan loss provisions/customer loans (LLP), impaired loans/equity (ILE), opposite distance to default (DD) and opposite Z-score. It-1 refers to the lagged one period of the dependent variables. HHI is the income diversification using Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income. LCD is Loans/Customer Deposits, ETA refers to Equity/Total Assets, CIR is Cost to Income Ratio. J-statistic checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. Linearity test checks whether threshold exists. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level

36

0.192

Upper regime 0.238

(0.018)

0.947

(0.108)

25.652***

(0.054)

(3.773)

27.831***

(0.013)

25.638***

(0.474)

(0.002)

−0.492*** 9.019**

(0.350)

−8.484**

0.610***

(0.222)

(2)

(0.002)

(1) 0.777***

(4)

−0.036*** −0.945*** −0.047*** −2.325*** −0.005

0.609

J-statistic

Threshold

CIR

ETA

(3)

(2)

(1)

LLE

LLE

ILP

ILP

Z-score

Idiosyncratic risk DD

D-SIBs Idiosyncratic risk

Financial distress

G-SIBs

Table 4.2 (continued)

4.4 Empirical Estimation and Results 105

106

4 The Impact of Income Diversification on Chinese Banks …

China’s Domestic Systemically Important Banks We now move to examine the efficiency effect of diversification for China’s domestic systemically important banks (D-SIBs). We can see that the empirical results for D-SIBs are similar to those for the whole sample. That is, with a lower level of diversification, the shift from traditional banking towards non-interest activities significantly increases the banks’ risk level in terms of both idiosyncratic risk and the probability of default. However, the results show that once D-SIBs have exceeded a certain threshold point, they benefit from a significant risk reduction derived from an increase in income diversity and a shift from interest to non-interest income. The results for both idiosyncratic risk and financial distress are consistent at the 1% significance level.

Other Chinese Banks The third column of Table 4.2 reports a different result compared with those for the other two sub-groups. Here again, it shows an inverse U-shape, where increased proportion of banks’ non-lending business will lead initially to an increase in both idiosyncratic risk and banks’ financial distress. However, after achieving a certain threshold point of diversification, N-SIBs will benefit from a reduction of risk. The difference in results across the three groups could reflect the different size of banks, from large G-SIBs to medium-sized D-SIBs and smaller N-SIBs. For G-SIBs, the process of diversification can make banks more stable regardless of the level of diversification. However, our results for D-SIBs and N-SIBs echo the work of Gamra and Plihon (2011), who use data from both East-Asian and Latin-American markets and find that income diversification makes banks less stable in the early stages of engaging in mixed business lines, but that they become more stable with the expansion of non-interest activity. These differences can be explained by the fact that mediumsized D-SIBs and small-sized N-SIBs are still in the early stages of diversification with lower overall diversification level, combined with less accumulation of specialist expertise, weak risk management and unsound internal regulatory systems, and are thus ill-equipped to deal with riskier non-traditional businesses.

4.4.3 Components of Non-interest Income and Risk Different components under the general heading of non-interest business may perform differently and, hence, may have different diversification effects (Stiroh and Rumble 2006; Chiorazzo et al. 2008; Roy 2015). Brunnermeier et al. (2012) suggest that trading and fee-based activities may have different risk implications, as trading activity is often accompanied by a greater increase in the variability in profits. In contrast, the fee-based activity is highly correlated with traditional business (Zhou 2014), which would help the banks obtain superior information. This prompts us to

4.4 Empirical Estimation and Results

107

explore the issue further by decomposing total operating income into three revenue classes and then investigate into the risk implications of each class. These components are fee and commission (Model 5), trading (Model 6), and other non-interest incomes (Model 7). Each of these three types of revenue is expressed as a share of total operating income. Regression results are reported in Table 4.3. Such positive effects on banks’ overall risk are mainly driven by the trading income. This result is in line with Estrella’s (2001) finding that acquisition of securities firms would cause highly volatile returns in the US market. According to Boot and Ratnovski (2012), trading activity allocates too much spare capital to trading expost and takes away too many resources from traditional business. Therefore, such resources misallocation towards to trading business would raise banks’ probability of failure. However, once a bank has exceeded the threshold level, all of those three activities will bring benefits due to risk reduction. The second, third and fourth columns of Table 4.3 show the regression results for G-SIBs, D-SIBs and N-SIBs respectively. For all three groups, the trading and other non-interest activities exhibit an inverse U-shaped relationship with bank risk. We find that the risk for N-SIBs is more sensitive to the increase in trading activities in the early stage. At the same time, this group has the lowest threshold for trading activities. Hence, although the expansion of trading activities might cause more serious problems for N-SIBs in terms of financial distress, such activities will eventually help the banks to reduce their level of risk. Fee-based activities exert different effects on banks’ overall risk level among the three groups. First, G-SIBs can always achieve benefits of risk reduction by increasing their fee-based activities. Moreover, when the proportion of fee and commissions achieves 13.06%, the coefficient changes from −0.082 to −0.148, indicating that the increase of such activity could play an accelerator role to reduce G-SIBs’ risk and make these banks more and more stable. However, the results for D-SIBs are consistent with those for the whole sample, showing an inverse U-shape with the threshold at 10.322%. The regression results for N-SIBs indicate that banks in this group cannot obtain benefits from the expansion of fee and commissions in either the lower or upper regime. The above result can be explained by the fact that G-SIBs have skills and expertise in fee-based activities, since most banks started getting involved in this kind of business in the 1990s, being followed later by D-SIBs and finally by N-SIBs. In the case of banks with well-established systems and skilled workers, the expansion of this business can provide continuing risk-reduction benefits. Meanwhile, for D-SIBs, the sign for fee-based income in the lower regime changes from negative to positive. Thus there exists an inverse U-shaped relation between fee-based income and banks’ risk indicator, with the threshold at 10.322%. Next, for N-SIBs, although the expansion of fee-based income cannot reduce banks’ risk in either the lower or upper regime, the coefficient changes from 0.593 with 1% significance level in the lower regime, to 0.294 with insignificant level in the upper regime. We suggest that this risk-enhancement result is mainly due to an internal problem whereby N-SIBs are more concerned with their interest business, while with regard to the non-traditional business they focus more on trading and

COM

It-1

Upper regime

ETA

LCD

LTA

OTH

TRAD

COM

It-1

Lower regime

0.056*

(0.019)

(0.054)

−0.004

(0.064)

0.119*

−0.148*** (0.041)

(0.877)

0.480

(0.21)

−0.461***

(0.290)

1.627***

(0.522)

(0.953)

−0.180

(0.37)

−1.779*** −0.982*** −0.263

(0.048)

−0.01

(0.033)

(0.118)

0.672

0.189

(0.21)

−0.277*

(0.154)

(0.444)

(0.042)

(0.036)

−0.605

−0.048

−0.085**

(0.418)

(0.090)

(0.100)

(0.051)

−0.080*

0.186**

−0.208*** −0.042

(1.227)

−1.287

(2.640)

−6.603**

(0.025)

−0.011

(0.355)

0.831***

2.545*** (1.009)

2.204**

(0.656)

(1.431)

(2.123)

−3.044

(7)

(1.009)

2.523***

(0.927)

−0.415

(6)

3.138**

(0.005)

(0.260)

(0.122)

(0.247) −0.082***

(0.343)

(0.285)

0.274

0.180

0.923***

−0.097

1.364***

G-SIBs (5)

(7)

(5)

(6)

Whole sample

(0.986)

2.158**

(0.669)

2.701***

(6)

(0.189)

(0.008)

0.024***

(0.171)

1.750***

(0.069)

1.174***

(7)

N-SIBs

(0.026)

(0.009)

−0.154***

(0.032)

0.619***

(0.023)

(0.958)

3.555***

(0.732)

−0.363*** 1.829**

(0.068)

(0.054)

1.562***

(0.162)

1.038***

(0.008)

(6)

(0.076)

0.056

(0.248)

0.593**

(0.902)

(0.231)

0.294

(0.737)

(0.019)

−0.037*

(0.014)

−0.108***

(0.611)

1.858***

(0.145)

1.779***

(7)

(0.229)

(0.142)

(continued)

(0.101)

1.394***

(0.107)

−1.256*** 0.056

(0.026)

2.535*** 1.487***

(0.302)

−0.203

(0.031)

0.036

(0.018)

−0.016

(1.761)

9.052***

(0.171)

2.834*** 1.907***

(5)

−0.247*** −0.129*** −0.216*** −0.036

(0.004)

−0.011*** 0.418**

(0.031)

0.021

(0.035)

0.982***

(5)

D-SIBs

Table 4.3 Income diversification and risk for 35 Chinese banks, G-SIBs, D-SIBs and N-SIBs from 2007 to 2016

108 4 The Impact of Income Diversification on Chinese Banks …

315

Observations 315

−0.044

0.551

315

0.000

0.369

0.820

(0.200)

0.633***

(1.165)

36

0.093

0.292

0.254

(0.174)

13.060***

(1.112)

−0.181

(0.107)

−0.012

(0.228)

0.241***

0.181***

(0.14)

0.204

(0.068)

0.018***

(0.003)

0.006**

(0.005)

36

0.032

0.504

0.373

(0.284)

0.670**

(0.428)

36

0.095

0.464

0.515

(0.076)

1.017***

(0.82)

81

0.051

0.364

0.158

(0.260)

10.322***

(0.057)

−1.380*** −2.174*** 0.290***

(0.054)

−0.007

(0.073)

D-SIBs (5)

81

0.623

0.205

0.382

(0.014)

1.404***

(1.349)

2.354*

(0.08)

0.160**

(0.071)

0.130*

(0.527)

−0.456

(6)

81

0.036

0.478

0.678

(0.052)

0.631***

(0.166)

0.260

(0.018)

0.007

(0.112)

0.418***

(0.013)

−0.053***

(7)

N-SIBs

198

0.062

0.403

0.762

(1.188)

(0.121)

−0.435***

(7)

(0.096)

−0.084

(0.024)

−0.009

(0.014)

198

0.079

0.698

0.220

(0.059)

198

0.000

0.289

0.172

(0.179)

0.701***

(0.225)

0.137

(0.037)

−0.028

(0.019)

−0.046*** −0.046**

(0.092)

−0.098

(6)

5.103*** 0.340***

(0.205)

0.54***

(0.039)

0.023

(0.054)

0.023

(5)

This table reports the first-differenced GMM threshold dynamic panel model developed by Seo and Shin (2016). Our dependent variable is insolvency risk using a accounting data based opposite Z-score. It-1 refers to the lagged one period of the dependent variables. COM is the ratio of net fee and commission income to total operating income; TRA is ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. LCD is Loans/Customer Deposits, ETA refers to Equity/Total Assets, CIR is Cost to Income Ratio. J-statistic checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. Linearity test checks whether threshold exists. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level

0.065

Linearity test 0.001

0.620

0.630

0.726

0.423

Upper regime

0.437

−0.319

(0.328)

−0.017

(0.213)

(0.426)

(0.025)

(0.037)

6.869***

−0.035

0.037

(0.563)

−0.014

(0.066)

(0.114)

−0.064

(0.048)

−0.164*** 0.025

(0.041)

−0.842 (0.879)

−1.325**

(0.665)

(0.145)

(7)

(0.550)

−0.340

(6)

−0.299**

J-statistic

Threshold

ETA

LCD

LTA

OTH

TRAD

G-SIBs (5)

(7)

(5)

(6)

Whole sample

Table 4.3 (continued)

4.4 Empirical Estimation and Results 109

110

4 The Impact of Income Diversification on Chinese Banks …

other activities rather than on fee-based income: this group has the lowest level of average HHI (13.39) among the three groups (25.12 for G-SIBs and 18.92 for D-SIBs). Moreover, among all non-interest activities they focus mainly on more profitable and high-leverage trading and other non-interest activities, whereas they do not have sufficient ability to deal with the risk that is generated from this kind of non-traditional activity.

4.4.4 Income Diversification and Different Types of Risk Having discussed effects of different components of non-traditional income on the overall risk of the banks, we now investigate how income diversification affects different types of risk. We follow Valverde and Fernández (2007) to decompose overall risk faced by Chinese banks into its three components, namely, credit, liquidity and interest rate risk. Table 4.4 presents the estimation results for the effects of diversification on each of these risks, i.e., on credit risk (Model 8), liquidity risk (Model 9) and interest rate risk (Model 10). The first column of Table 4.4 reports the results for the whole sample; it shows the regression outcome when employing the proxy of credit risk as the dependent variable. We find that income diversification can always make banks riskier in terms of credit risk (CREDIT). Next, the results for the liquidity risk shown in Model 9 differ from those for credit risk, indicating that the liquidity risk (LIQUIDITY) is significantly negatively correlated with income diversification in both the lower and upper diversification regimes. Finally, Model 10 under the whole sample presents the estimation results in relation to interest rate risk (INTEREST) when it is specified as the dependent variable. Turning to the next three columns, for G-SIBs, D-SIBs and N-SIBs respectively, we find a consistent result in terms of the liquidity risk, which implies that diversification brings benefits to banks’ liquidity risk for both diversified and less diversified banks. This may be because, compared with traditional lending business, non-interest activities are less likely to be affected by the regulator’s capital restrictions, such as the capital adequacy requirement. Rather, an increase in non-interest activities will provide liquidity for banks and, therefore, reduce bank liquidity risk. However, the regression results for credit and interest rate risks vary among the three groups. In detail, for both G-SIBs and D-SIBs, we see that there is an inversion of the signs for the diversification effect on credit risk between the low and high diversification regimes. At lower levels of diversification, income diversification will generate more credit risk. Once a bank has passed the threshold level, the increase of non-interest income will then lead to a reduction of credit risk. However, 59.7% of observations for G-SIBs, and 85.5% for D-SIBs, are in the lower diversification regime. Therefore, currently, most Chinese systemically important banks are suffering an increase in credit risk as a result of their relatively low level of diversification.

LCD

LTA

HHI

It-1

Upper regime

ETA

LCD

LTA

HHI

It-1

Lower regime

0.045

(0.030)

−0.109***

−0.685**

(0.337)

−0.204

(0.603)

0.100***

(0.024)

0.064*

(0.038)

(0.587)

3.540**

(1.682)

−0.399**

(0.200)

−0.047***

−1.072**

(0.451)

0.057***

(0.01)

Model 9

(0.073)

−0.913***

(0.085)

−0.198**

(0.096)

−1.360***

(0.056)

(0.481)

(0.182)

−0.040*** 0.038

0.747

0.007 (0.028)

−0.059***

0.039**

0.225*** (0.013)

−0.066***

(0.336)

(0.01)

(0.004)

−0.030*** −0.399

(0.019)

−0.214*** −0.251***

(0.083)

−0.719*** −1.927***

(0.177)

−2.325*** 0.806*

(0.016)

0.133***

(0.028)

0.308***

(0.004)

0.038***

(0.200)

−1.370*** 0.399***

Model 8

D-SIBs

−0.039***

(0.005)

0.125***

(0.004)

−0.108***

(0.021)

−0.056***

(0.207)

−0.500**

(0.019)

−0.028

(0.033)

0.004

(0.006)

−0.031***

(0.237)

0.883***

Model 10

(0.113)

−1.450***

(0.173)

0.258

(0.14)

−0.627***

(0.047)

−0.095**

Model 9

(0.701)

(0.059)

−1.234***

(0.036)

0.009

(0.003)

0.595***

(0.069)

−0.019*** 0.497***

(0.003)

−0.006**

(0.170)

−0.539*** −0.103***

(0.051)

N-SIBs

0.038

(0.01)

0.090***

(0.032)

−0.077**

(0.516)

0.574

(0.151)

−0.340**

(0.012)

0.042***

(0.017)

0.187***

(0.035)

−0.245***

(0.994)

−1.229

Model 10

(1.05)

−0.769

(0.906)

−3.201***

(0.410)

(0.62)

0.082***

(0.006)

0.022***

(0.004)

0.032***

(0.168)

−−0.322*

(0.062)

−1.054***

(0.431)

−0.907**

(0.575)

−0.320

(0.239)

0.501**

(1.937)

−0.245*** 2.270

(0.010)

−0.031*** −0.198

(0.011)

0.076***

(0.022)

0.102***

(0.187)

(continued)

−0.034***

(0.006)

−0.042***

(0.007)

−0.044***

(0.09)

0.252***

(0.447)

0.756*

(0.108)

0.017

(0.105)

−0.214**

(0.437)

0.305

(3.15)

0.676

Model 10

LIQUIDITY INTEREST Model 9

−1.096*** 1.532***

Model 8

LIQUIDITY INTEREST CREDIT

−0.266*** 3.386***

(0.002)

0.020***

(0.003)

−0.005

(0.005)

0.034***

(0.143)

0.319**

Model 8

LIQUIDITY INTEREST CREDIT

(0.006)

(0.011)

(0.141)

−0.230

(0.324)

−0.072

(0.257)

−0.146

(0.051)

0.03

−0.059***

(0.013)

−0.063*** −1.742***

(0.02)

(0.015)

(0.189)

1.370***

(0.205)

1.548**

(0.733)

0.595***

Model 10

Model 9

Model 8

G-SIBs

LIQUIDITY INTEREST CREDIT

CREDIT

Whole sample

Table 4.4 Income diversification and risk for 35 Chinese banks from 2007 to 2016

4.4 Empirical Estimation and Results 111

0.095

Linearity test

0.229

315

0.004

0.269

17.442***

315

0.084

0.403

1

(0.671)

25.922***

36

0.000

0.403

0.971

(0.047)

36

0.046

0.344

0.986

(0.061)

26.939***

(0.104)

−0.448*** −3.624*** (0.019)

−0.177**

(0.031)

(0.001)

Model 9

(0.013)

(0.073)

D-SIBs

36

0.000

0.442

0.986

(0.089)

25.658***

(0.024)

−0.563***

(0.001)

Model 10

81

0.033

0.145

0.702

(0.334)

26.913***

(0.042)

0.210***

(0.005)

Model 8

LIQUIDITY INTEREST CREDIT

N-SIBs

81

0.007

0.343

0.995

(0.324)

21.098***

(2.231)

−8.589***

(0.115)

Model 9

81

0.067

0.188

0.521

(1.549)

25.682***

(0.284)

0.002

(0.032)

Model 10 (0.32)

198

0.000

0.376

0.452

(0.660)

14.579***

(0.042)

198

0.000

0.466

1

(0.929)

12.961***

(2.062)

198

0.097

0.873

0.821

(0.455)

6.265***

(0.121)

−0.298**

(0.009)

Model 10

LIQUIDITY INTEREST Model 9

−0.218*** 6.859***

(0.011)

Model 8

LIQUIDITY INTEREST CREDIT

This table reports the first-differenced GMM threshold dynamic panel model developed by Seo and Shin (2016). Our dependent variables are loan default/total loans (CREDIT), liquidity assets/short term funding (LIQUIDITY), and interest rate risk calculated by interbank rate - interest rate for customer deposits; (INTEREST). It-1 refers to the lagged one period of the dependent variables. HHI indicates the income diversification using Herfindahl-Hirschman Index with the components of interest income and three activities under non-interest income. LCD is Loans/Customer Deposits, ETA refers to Equity/Total Assets, CIR is Cost to Income Ratio. J-statistic checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. Linearity test checks whether threshold exists. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level

Observations 315

0.102

0.505

Upper regime

20.560***

(1.953)

14.996***

(0.813)

0.29

(1.96)

0.03

(0.067)

J-statistic

Threshold

ETA

(1.289)

Model 8

(0.016)

Model 10

Model 9

Model 8

G-SIBs

LIQUIDITY INTEREST CREDIT

CREDIT

Whole sample

Table 4.4 (continued)

112 4 The Impact of Income Diversification on Chinese Banks …

4.4 Empirical Estimation and Results

113

However, the result for N-SIBs is different. In both the lower and upper regimes, the expansion of non-interest activities does not reduce banks’ credit risk at all. This result is also consistent with our previous results in Sect. 4.2, as the credit risk is mainly created from traditional lending and fee-based activities, which are highly correlated with the interest income. Such low ability to deal with fee-based income and its risks leads to the enhancement of credit risk from income diversification. With regard to interest risk, for G-SIBs and D-SIBs we find a consistent result, where similar to the situation with liquidity risk, in both the lower and upper regimes diversification can reduce interest risk. One plausible explanation for this is that banks’ exposure to changes in interest rates is the result of bank asset-liability management. A change in the interest rate would directly affect banks’ interest income and their liabilities; in particular, an increase in the interest rate would largely reduce the interest margin and the banks’ equity capital. Therefore, by engaging in non-interest activities banks can reduce their exposure to interest rate risk, especially when compared with banks that have a large percentage of mortgage lending business (Delong 2001). For N-SIBs, which tend to be less diversified, an increase in the diversification level can increase the banks’ exposure to interest rate risk. Then, after passing a low threshold level (6.265), this correlation becomes negative. We suggest that this result is because the overall diversification level for N-SIBs is still relatively low, where the proportion of non-interest income occupies only 7.29% of total operating income. Hence, in the early stages of diversification, the increase of non-interest income will not have a big effect on banks’ overall interest risk.

4.5 Conclusion This chapter examines to what extent income diversification affects the risks of Chinese banks. The majority of previous studies indicate a linear relation between income diversification and risk, finding either a negative correlation that suggests banks should diversify, or a positive correlation that suggests banks should remain focused on core business. However, by adopting the first-differenced GMM estimator for the dynamic threshold panel data model, we get results showing that in the case of Chinese systemically important banks, the relation between income diversification and risk is not linear, and the effects of diversification vary with different sources of non-interest activities and different measures of risk. Generally, there exists an inverse U-shaped relation between diversification level and risk in the Chinese banking industry. Income diversification will reduce bank risk only after the bank has passed a certain threshold of income diversification. This pattern of relationship seems to be driven mainly by the learn-by-doing effect and the mitigation of agency problems, which result from the expansion of noninterest activities. After dividing the sample into three groups based on the BIS definition and the Chinese regulator’s classification, we find that, for G-SIBs, diversification has a significant negative effect on both banks’ idiosyncratic risk and banks’

114

4 The Impact of Income Diversification on Chinese Banks …

financial distress. However, for D-SIBs and N-SIBs, the effects again exhibit an inverse U-shape, where in the early stages of income diversification banks incur a discount and become less stable, while they become more stable after achieving a certain threshold point of diversification. We suggest that these differences are due to different diversification strategies and risk preferences, which lead to different income structures. The diversification effect on bank risk turns out to be not uniform across different business lines. Where there is only a low level of diversification, both trading and other non-interest activities will lead to increased exposure to risk; only once the bank has achieved a certain diversification level will these activities start to provide diversification benefits that will lower bank risk. With regard to the different components of non-interest business, fee and commissions activities will decrease bank risk for G-SIBs and increase risk for N-SIBs regardless of diversification level, while for DSIBs there is an inverse result, positive for banks with a lower level of diversification and negative for more diversified banks. We also examine the diversification effect across different types of risk. Evidence suggests that for G-SIBs and D-SIBs diversification can reduce credit risks only when the banks have passed a certain threshold point, while for N-SIBs diversification cannot bring any improvement for credit risk regardless of diversification level. Similarly, the interest rate risk will be reduced only for highly diversified GSIBs and D-SIBs. Finally, diversification will always reduce the liquidity risk, for all three groups. However, the reduction of the liquidity risk cannot fully offset the enhancement of the credit and interest rate risks. Therefore, it is necessary for banks to cross the threshold in order to obtain the risk-reduction benefits from diversification. Furthermore, this implies that to fully reap the benefits of risk reduction from income diversification, banks need to accumulate sufficient banking human capital and to establish an effective supervision system.

References Acharya VV, Yorulmazer T (2008) Information contagion and bank herding. J Money Credit Bank 40(1):215–231 Acharya V, Hasan I, Saunders A (2006) Should banks be diversified? evidence from individual bank loan portfolios. J Bus 79(3):1355–1412 Angkinand A, Wihlborg C (2010) Deposit insurance coverage, ownership, and banks’ risk-taking in emerging markets. J Int Money Financ 29(2):252–274 Arellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev Econ Stud 58(2):277–297 Ashraf D, Ramady M, Albinali K (2016) Financial fragility of banks, ownership structure and income diversification: empirical evidence from the GCC region. Res Int Bus Financ 38:56–68 Baele L, De Jonghe O, Vander Vennet R (2007) Does the stock market value bank diversification? J Bank Financ 31(7):1999–2023 Barry TA, Lepetit L, Tarazi A (2011) Ownership structure and risk in publicly held and privately owned banks. J Bank Financ 35(5):1327–1340

References

115

Berger AN, Hasan I, Zhou M (2009) Bank ownership and efficiency in China: what will happen in the world’s largest nation? J Bank Financ 33(1):113–130 Berger AN, Hasan I, Zhou M (2010a) The effects of focus versus diversification on bank performance: evidence from Chinese banks. J Bank Financ 34(7):1417–1435 Berger AN, Hasan I, Korhonen I, Zhou M (2010b) Does diversification increase or decrease bank risk and performance? evidence on diversification and the risk-return trade-off in banking. BOFIT discussion paper No. 9/2010 Berger PG, Ofek E (1995) Diversification’s effect on firm value. J Financ Econ 37(1):39–65 Bharath ST, Shumway T (2008) Forecasting default with the Merton distance to default model. Rev Financ Stud 21(3):1339–1369 Bonin JP, Hasan I, Wachtel P (2005) Privatization matters: bank efficiency in transition countries. J Bank Financ 29(8–9):2155–2178 Boot AW, Ratnovski L (2012) Post-crisis banking regulation: evolution of economic thinking as it happened on Vox. CEPR Press, London Brei M, Yang X (2015) The universal bank model: Synergy or vulnerability?. Université de Paris Ouest Nanterre La Défense No. hal-01671512 Brunnermeier M, Dong GN, Palia D (2012) Banks’ non-interest income and systemic risk. AFA 2012 Chicago Meetings No. 25 Busch R, Kick T (2009) Income diversification in the German banking industry. Bundesbank Series 2 Discussion Paper No. 09/2009 Campa JM, Kedia S (2002) Explaining the diversification discount. J Financ 57(4):1731–1762 Campbell JY, Hilscher J, Szilagyi J (2008) In search of distress risk. J Financ 63(6):2899–2939 Caner M, Hansen BE (2004) Instrumental variable estimation of a threshold model. Econom Theory 20(5):813–843 Chen CR, Huang YS, Zhang T (2017) Non-interest income, trading, and bank risk. J Financ Serv Res 51(1):19–53 Chen PF, Zeng JH (2014) Asymmetric effects of households’ financial participation on banking diversification. J Financ Stab 13:18–29 Chen X, Wang X, Wu DD (2010) Credit risk measurement and early warning of SMEs: an empirical study of listed SMEs in China. Decis Support Syst 49(3):301–310 Chiorazzo V, Milani C, Salvini F (2008) Income diversification and bank performance: evidence from Italian banks. J Financ Serv Res 33(3):181–203 Cornett MM, Guo L, Khaksari S, Tehranian H (2010) The impact of state ownership on performance differences in privately-owned versus state-owned banks: an international comparison. J Financ Intermed 19(1):74–94 De Jonghe O, Diepstraten M, Schepens G (2015) Banks’ size, scope and systemic risk: what role for conflicts of interest? J Bank Financ 61(1):S3–S13 DeLong GL (2001) Stockholder gains from focusing versus diversifying bank mergers. J Financ Econ 59(2):221–252 Demirgüç-Kunt A, Huizinga H (2010) Bank activity and funding strategies: the impact on risk and returns. J Financ Econ 98(3):626–650 Demsetz RS, Strahan PE (1997) Diversification, size, and risk at bank holding companies. J Money Credit Bank 300–313 Deng X, Li S (2006) An empirical study of income diversification and performance: evidence from Chinese commercial bank. J Spec Zone Econ 9:339–341 (in Chinese) DeYoung R, Rice T (2004) Noninterest income and financial performance at US commercial banks. Financ Rev 39(1):101–127 DeYoung R, Roland KP (2001) Product mix and earnings volatility at commercial banks: evidence from a degree of total leverage model. J Financ Intermed 10(1):54–84 DeYoung R, Torna G (2013) Nontraditional banking activities and bank failures during the financial crisis. J Financ Intermed 22(3):397–421 Doumpos M, Gaganis C, Pasiouras F (2016) Bank diversification and overall financial strength: international evidence. Financ Mark Inst Instrum 25(3):169–213

116

4 The Impact of Income Diversification on Chinese Banks …

Elsas R, Hackethal A, Holzhäuser M (2010) The anatomy of bank diversification. J Bank Financ 34(6):1274–1287 Elyasiani E, Wang Y (2008) Non-interest income diversification and information asymmetry of bank holding companies. Unpublished manuscript Estrella A (2001) Mixing and matching: prospective financial sector mergers and market valuation. J Bank Financ 25(12):2367–2392 Gambacorta L, Scatigna M, Yang J (2014) Diversification and bank profitability: a nonlinear approach. Appl Econ Lett 21(6):438–441 Gamra SB, Plihon D (2011) Revenue diversification in emerging market banks: implications for financial performance. Unpublished manuscript Goddard J, McKillop D, Wilson JO (2008) The diversification and financial performance of US credit unions. J Bank Financ 32(9):1836–1849 Gurbuz AO, Yanik S, Ayturk Y (2013) Income diversification and bank performance: evidence from Turkish banking sector. J BRSA Bank Financ Mark 7(1):9–29 Hansen BE (1999) Threshold effects in non-dynamic panels: estimation, testing, and inference. J Econom 93(2):345–368 Hansen BE (2000) Sample splitting and threshold estimation. Econometrica 68(3):575–603 Hayden E, Porath D, Westernhagen NV (2007) Does diversification improve the performance of German banks? evidence from individual bank loan portfolios. J Financ Serv Res 32(3):123–140 Hovakimian A, Kane EJ, Laeven L (2012) Tracking variation in systemic risk at us banks during 1974–2013. In: National Bureau of economic research working paper No. w18043 Hsiao C, Zhang J (2015) IV, GMM or likelihood approach to estimate dynamic panel models when either N or T or both are large. J Econom 187(1):312–322 Hsieh MF, Chen PF, Lee CC, Yang SJ (2013) How does diversification impact bank stability? the role of globalization, regulations, and governance environments. Asia-Pac J Financ Stud 42(5):813–844 Iskandar-Datta M, McLaughlin R (2007) Global diversification: evidence from corporate operating performance. Corp Ownersh Control 4(4):228–242 Kobayashi T (2012) Diversity among banks may increase systemic risk. Kobe University Discussion Paper No. 1213 Köhler M (2014) Does non-interest income make banks more risky? retail-versus investmentoriented banks. Rev Financ Econ 23(4):182–193 Köhler M (2015) Which banks are more risky? the impact of business models on bank stability. J Financ Stab 16:195–212 Kourtellos A, Stengos T, Tan CM (2016) Structural threshold regression. Econom Theory 32(4):827–860 Laeven L, Levine R (2009) Bank governance, regulation and risk taking. J Financ Econ 93:259–275 Laeven L, Levine R (2007) Is there a diversification discount in financial conglomerates? J Financ Econ 85(2):331–367 Lee CC, Hsieh MF, Yang SJ (2014) The relationship between revenue diversification and bank performance: do financial structures and financial reforms matter? Jpn World Econ 29:18–35 Lepetit L, Nys E, Rous P, Tarazi A (2008) Bank income structure and risk: an empirical analysis of European banks. J Bank Financ 32:1452–1467 Li L, Zhang Y (2013) Are there diversification benefits of increasing noninterest income in the Chinese banking industry? J Empir Financ 24:151–165 Martens M, Kofman P, Vorst TC (1998) A threshold error-correction model for intraday futures and index returns. J Appl Econom 13(3):245–263 Mercieca S, Schaeck K, Wolfe S (2007) Small European banks: benefits from diversification? J Bank Financ 31(7):1975–1998 Mergaerts F, Vander VR (2016) Business models and bank performance: a long-term perspective. J Financ Stab 22:57–75 Meslier C, Morgan DP, Samolyk K, Tarazi A (2014) The Benefits of intrastate and interstate geographic diversification in banking. Unpublished manuscript

References

117

Nguyen J (2012) The relationship between net interest margin and noninterest income using a system estimation approach. J Bank Financ 36(9):2429–2437 Nguyen M, Perera S, Skully M (2016) Bank market power, ownership, regional presence and revenue diversification: evidence from Africa. Emerg Mark Rev 27:36–62 Pennathur AK, Subrahmanyam V, Vishwasrao S (2012) Income diversification and risk: does ownership matter? an empirical examination of Indian banks. J Bank Financ 36:2203–2215 Ramírez-Rondán N (2015) Maximum likelihood estimation of dynamic panel threshold models. In: Peruvian economic association working paper No. 2015–32 Roy TS (2015) Banking innovations and new income streams: Impact on banks’ performance. University Library of Munich. Working paper Sanya S, Wolfe S (2011) Can banks in emerging economies benefit from revenue diversification? J Financ Serv Res 40(1–2):79–101 Seo MH, Linton O (2007) A smoothed least squares estimator for threshold regression models. J Econom 141(2):704–735 Seo MH, Shin Y (2016) Dynamic panels with threshold effect and endogeneity. J Econom 195(2):169–186 Stein JC (2002) Information production and capital allocation: decentralized versus hierarchical firms. J Financ 57(5):1891–1921 Stiroh KJ (2004) Diversification in banking: is noninterest income the answer? J Money Credit Bank 36(5):853–882 Stiroh KJ, Rumble A (2006) The dark side of diversification: the case of US financial holding companies. J Bank Financ 30(8):2131–2161 Tabak BM, Fazio DM, Cajueiro DO (2011) The effects of loan portfolio concentration on Brazilian banks’ return and risk. J Bank Financ 35(11):3065–3076 Tiao GC, Tsay RS (1994) Some advances in non-linear and adaptive modelling in time-series. J Forecast 13(2):109–131 Tong H (1978) On a threshold model. In: Pattern recognition and signal processing, pp 101–141 Vallascas F, Keasey K (2012) Bank resilience to systemic shocks and the stability of banking systems: small is beautiful. J Int Money Financ 31(6):1745–1776 Valverde SC, Fernández FR (2007) The determinants of bank margins in European banking. J Bank Financ 31(7):2043–2063 Vassalou M, Xing Y (2004) Default risk in equity returns. J Financ 59(2):831–868 Villalonga, B. (2004). Does diversification cause the “diversification discount”?. Financ Manag 5–27 Wagner W (2010) Diversification at financial institutions and systemic crises. J Financ Intermed 19(3):373–386 Yu P (2012) Likelihood estimation and inference in threshold regression. J Econom 167(1):274–294 Zhou K (2014) The effect of income diversification on bank risk: evidence from China. Emerg Mark Financ Trade 50(sup3):201–213 (in Chinese)

Chapter 5

The Impact of Income Diversification on Chinese Banks: Bank Efficiency

Abstract This chapter examines the efficiency of the Chinese banking sector in relation to banks’ diversification operations. In a two-step approach, the chapter first calculates both cost and profit efficiency scores via stochastic frontier analysis using the method of within maximum likelihood estimation (WMLE). Then, the investigation employs a dynamic Tobit model in order to shed light on the diversification-efficiency nexus in the Chinese banking market.

Chapters 3 and 4 have explored the benefits and risks associated with Chinese banks’ engagement in income diversification. Another critical aspect of the effects of banks’ income diversification is the extent to which banks’ efficiency would be affected. Therefore, in order to provide a fuller picture of the consequences of banks’ diversification, this chapter is devoted to investigating the efficiency of the diversifying banks in China. In a two-step approach, first, bank efficiency scores, both cost and profit, are calculated for the three main categories of Chinese banks. The computation is based on stochastic frontier analysis using within maximum likelihood estimation (WMLE). In the second step, a dynamic Tobit model is estimated to examine unobserved, time-invariant bank heterogeneity. Finally, the empirical findings are discussed with reference to the different banking groups.

5.1 Introduction The impact of diversification on bank efficiency has been a contentious issue in the literature. One school in the debate subscribes to the view that diversification is efficiency enhancing. Based on the portfolio theory, Meng et al. (2017) argue that diversification can yield separation of risks and hence is beneficial to banks’ efficiency. Additional channels through which diversification may enhance banks’ efficiency include gains accruing due to scale and scope economies (Meslier et al. 2014), tax reduction as a result of higher financial leverage (Elsas et al. 2010),

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0_5

119

120

5 The Impact of Income Diversification on Chinese Banks …

improved corporate governance (Lin et al. 2012), and a reduction in asymmetric information between borrowers and lenders (Akhigbe and Stevenson 2010). On the other hand, the opposing school argues that shifting from focused to diversified operations has a negative impact on bank efficiency. This may be caused by, for example, the increased correlation of internal businesses and the growing complexity across business lines (Elyasiani and Wang 2012), where the presence of too many business lines may lead to disorganized management. Hence, more diversified banks can incur additional overhead costs (Elsas et al. 2010), inefficient cross-subsidization (Klein and Saidenberg 2010), inefficient internal capital markets within multinational groups (Curi et al. 2015), and the problems related to ‘too big to fail’ status (Quaglia and Spendzharova 2017). Furthermore, diversification can cause banks to have no power or incentive to monitor (Allen et al. 2011; Adzobu et al. 2017). Ultimately, whether or not diversification can increase or reduce bank efficiency is a matter of empirical evidence. This chapter contributes to the debate through empirical investigation into the efficiency effect of diversification in the Chinese banking sector. As in previous chapters, the sample comprises 40 major Chinese commercial banks, which are divided into three sub-groups: Global Systemically Important Banks (G-SIBs), Domestic Systemically Important Banks (D-SIBs), and banks that are not classified by the authorities as systemically important (N-SIBs). This grouping enables us to examine the possible heterogeneity of efficiency effects due to bank diversification. Similar research in the previous literature is mostly based on analysis of financial ratios, and employs balance sheet and stock market data (e.g. Stiroh and Rumble 2006; Demirgüç-Kunt and Huizinga 2010; Meslier et al. 2014). However, financial ratios suffer from several accounting biases, such that they are unlikely to offer a full account of the business mix and input prices (Titova 2016) and fail to provide information about managerial actions (Yeboah and Asirifi 2016) or the quality of service under complex business networks (LaPlante 2015). We deploy a two-stage approach in which the first step is to obtain the efficiency scores through the stochastic frontier analysis. Based on within maximum likelihood estimation (WMLE), as developed by Chen et al. (2014), stochastic frontier analysis (SFA) contains more information and more factors, which are difficult to quantify when using financial ratio based estimation and non-parametric approaches, such as data envelopment analysis (DEA). Furthermore, the SFA method employed by this research is particularly appropriate for the Chinese context, since issues regarding data availability mean that the panel data are relatively short. My approach to SAF analysis, which employs the first-difference data transformation, eliminates nuisance parameters, thus solving the incidental parameters problem and making the estimation of the efficiency scores unbiased (Greene 2005). In the next stage, we deploy the dynamic Tobit model to determine to what extent diversification drives Chinese banks’ efficiency. Unlike OLS and other linear estimation methods, the Tobit model is able to take into account the censored nature of the dependent variable. Since in the model in this study the efficiency scores are limited to between 0 and 1, the Tobit model is suitable for use in the regression on these

5.1 Introduction

121

variables and should yield consistent estimates that avoid the problems of bias and inconsistency associated with OLS (Souza and Gomes 2015). The dependent variable used in the regression is assumed to be half normally distributed; hence, the regression errors are subject to normal distribution. Most non-censored estimates, such as OLS, assume that the dependent variable can take on every negative or positive real number. This means ignoring fractionality, which may lead to biased estimation and inconsistent parameters (Greene 1980). In addition, other non-linear estimations, such as the fractional Probit model, are less likely to capture the dynamic effect. The presence of lagged dependent variables requires specification of the distribution of unobserved effects in a maximum likelihood framework, which leads to inconsistent results in the fractional Probit model (Papke and Wooldridge 2008). We apply the dynamic doubly censored Tobit model with a left censored bound of zero and a right censored bound of one to regress efficiency scores against banks’ income diversification (Elsas and Florysiak 2015). Several control variables are introduced into the model. This modelling choice allows for bank heterogeneity, does not require balanced panel data and is robust to missing data in unbalanced panels. Our study is also the first to investigate the diversification effect on banks’ efficiency across banking groups. This yields a number of insights, as banks in the three groups exhibit very different characteristics in terms of capital restriction, size and diversification level, which could help to shed light on, for example, whether the diversification effect varies with bank scale. The remainder of the chapter is organized as follows. Section 5.2 provides an overview of the relevant literature regarding bank efficiency and income diversification. The variables, data, and methodology are discussed in Sect. 5.3. Section 5.4 outlines the model specification and reports the empirical results. The conclusion is presented in Sect. 5.5.

5.2 Literature Review Two main theoretical views regarding the effect of diversification on banks’ efficiency have emerged in the literature, namely, the bank-based view and the market-based view. Built on the theory of financial intermediation, the bank-based view maintains that the benefits generated by diversification stem mainly from the information advantages (Bencivenga and Smith 1991; Saunders and Walter 1994; Lepetit et al. 2008; Elyasiani and Wang 2012). According to Elyasiani and Wang (2012), information can be treated as an important input factor to influence banks’ efficiency in terms of customer relation consolidation and credit screening. Alternatively, with mixed business lines, managers are more likely to lower their personal risk by over-diversifying their banks’ portfolios (Deng and Elyasiani 2008). As a result, non-interest income could result in excessive profits and bonuses to managers themselves. In this situation, banks would continually expand the range of non-traditional business, even if diversification of activities lowers the market valuation of the banking conglomerate (Demirgüç-Kunt and Huizinga 2010).

122

5 The Impact of Income Diversification on Chinese Banks …

While banks can reap diversification benefits through operational synergies (Lin et al. 2012), the attainment of these synergies relies on scale economies (Stiroh 2000; Sanya and Wolfe 2011). As argued by Drucker and Puri (2008) and Klein and Saidenberg (2010), the expansion of non-interest business is largely based on banks’ infrastructure, and a mixed business line strategy could help banks spread fixed costs and managerial overheads. Scale economies may also lead to operational benefits because cross-selling strategy companies could share the costs of monitoring, advertising and account maintenance, thus further reducing cost and hence improving banks’ production efficiency (Elyasiani and Wang 2012). From the perspective of resource allocation, whereas focused-business companies can only configure resources through external capital markets, diversified companies can be more effective through the internal capital market (Elyasiani et al. 2016). This could help banks to reallocate internal resources from less-profitable sectors to more-effective sectors. The market-based view introduces a competitive framework to explain how income diversification could result in efficiency improvement or discount. Several studies suggest that the diversification process could lead to more intense competition in the banking industry (e.g. Schaeck and Cihak 2010; Ariss 2010; Ibragimov et al. 2011; Beck et al. 2013). Authors such as Lepetit et al. (2008), Schaeck and Cihak (2010), Allen et al. (2011), Dell’Ariccia et al. (2012) and Amidu and Wolfe (2013) then believe that a more intense competitive environment could help banks achieve more efficient management of operation and risk. Conversely, some researchers argue that fiercer competition could exacerbate risk. Vives (2011) suggests two channels for such risk enhancement. First, increased interbank competition could exacerbate the coordination problem and hence increase banks’ frangibility. Second, managers’ risk-taking behaviour may change in the face of a more competitive environment. With more pressure on profits, there would be an increased incentive for managers to take on more excessive risk on either side of the bank’s balance sheet, resulting in higher fragility (Amidu and Wolfe 2013). In addition, in a more-competitive banking market banks earn less informational rent from their relationship with borrowers and would therefore have less incentive to properly screen borrowers, which would further increase the risk of fragility (Allen and Gale 2004; Beck et al. 2013). In empirical studies, bank efficiency can be evaluated through ratio analysis that employs accounting data (Cornett et al. 2006; Xu 2011; Jiang et al. 2013; Sun et al. 2013). However, accounting ratios are unlikely to offer a full account of the business mix and input prices (Havranek et al. 2016) because accounting-based efficiency measurements implicitly assume that all assets are equally costly to produce and that all locations have equal operation expenditures. Alternatively, data envelopment analysis (DEA) has been widely used in the literature for estimating bank efficiency (i.e. Biener et al. 2016; Aggelopoulos and Georgopoulos 2017). DEA is a non-parametric tool for performance evaluation and benchmarking. The method empirically measures efficiency of decision-making units by estimating the efficiency frontiers, and can also be used for benchmarking in operations management. Since its introduction by Charnes et al. (1978), DEA has been

5.2 Literature Review

123

applied to many industries, including the banking sector. In DEA, the efficiency is measured against the highest observed performance instead of an average (Hjalmarsson and Veiderpass 1992). As a non-parametric approach to measuring efficiency, the major advantage of DEA is that it does not assume a particular functional form/shape for the frontier; hence, it can be used when conventional cost and profit cannot be justified (Berger and Humphrey 1997). However, its drawbacks include a lack of measurement of errors and luck factors, sensitivity to outliers, an inability to measure absolute efficiency, and ignoring price information (Berger and Mester 1997; Fiorentino et al. 2006). More recently, researchers have increasingly used stochastic frontier analysis (SFA) to evaluate banks’ efficiency. SFA is a fully parameterized model and is now the mainstream parametric technique for efficiency analysis. Compared with nonparametric methods, such as DEA, stochastic methods could be more effective and robust when treating noisy data (LaPlante 2015). Due to consideration of random errors in the functional form, they can measure some factors that are very difficult to quantify, such as companies’ luck or even the influence of the weather. Originally developed by Aigner et al. (1977) and Meeusen and Van den Broeck (1977), SFA has been applied in a large number of efficiency studies, including studies of banks; see Berger and Humphrey (1997) for an extensive survey of 130 studies regarding the efficiency in financial institutions. Many varieties of the stochastic frontier model have appeared in the literature; see a major survey by Kumbhakar and Lovell (2003) and also Bauer (1990) and Greene (2008). The method involves estimation of both cost and profit efficiency (Kumbhakar and Lovell 2003). With regard to cost efficiency, SFA measures how far a firm is from full-cost minimization. In the profit efficiency analysis, the firm in question is treated as a profit-maximizer. Since the concept of profit efficiency takes into consideration the effects of output choice on costs and revenues, it is broader than that of cost efficiency. The empirical evidence regarding the effects of bank income diversification on efficiency has been mixed. Pasiouras et al. (2007) suggests that there is a positive effect on banks’ efficiency level. Such improvement in banks’ efficiency could be explained by the soft information provided through prolonged close interaction with clients (Gourlay et al. 2006). Doan et al. (2017) focuses on a cross-country case and suggests that banks can achieve significant profit efficiency gains through greater diversification of their risk exposure. In addition, a better mix of financial products also protects banks’ firm-specific human capital, as diversification through mergers and acquisition will help banks to expand the skill set of managers, which will in turn contribute to improving risk management and smoothing credit operations. Beccalli and Frantz (2009) employ SFA to generate estimates of cost and alternative profit efficiencies from 1991 to 2005. They find a marked improvement in cost efficiency and argue that this improvement would not have occurred in the absence of increases in both on- and off-balance sheet activities in EU mergers and acquisitions. Similar results are obtained by Doan et al. (2017), who, by adopting cross-country data, demonstrate that increased diversification tends to improve bank efficiency.

124

5 The Impact of Income Diversification on Chinese Banks …

Among the research on emerging markets, Kwan (2006) analyses the cost efficiency of Hong Kong banks; by employing SFA, he verifies that efficiency improvement is mainly generated from scale economies because of the increased contribution of non-interest income. Alhassan (2015) uses the same parametric approach to estimate both cost and profit efficiency of banks in Ghana, and detects efficiency improvement. However, some studies provide evidence for a diversification discount. The results of Wu’s (2008) investigation of 11 Taiwan banks suggest that an expansion of broad financial product lines would not bring an efficiency improvement but would further increase the losses from bad loans and erode banks’ original efficiency level. Similarly, Abbott et al. (2013), in their study of Australian banks over the period 1983– 2001, observe that an increase in the number of business lines through diversified growth does not drive significant improvement in banks’ overall efficiency. In particular, banks in the very early stages of diversification and business expansion through mergers with other financial institutions might not achieve competitive advantages in the short term as they lack relevant competence. Several studies support this view (e.g. Barth et al. 2013; Alhassan and Tetteh 2017), confirming that for banks, especially in the early stages of diversification, lack of experience causes inefficiency. This would be especially so in the case of emerging countries. Research has also found that bank size is relevant to the effect of income diversification on efficiency. Empirical evidence from the US and Europe markets, which accounts for the bulk of the current research in the field, generally suggests that an increase in non-interest activities can improve efficiency through economies of scale (Drake et al. 2009; Elsas et al. 2010), which implies that the benefits of income diversification are driven by a larger bank size. In contrast, small banks emphasize basic banking activities with low-cost funds and high-quality investments, thus limiting their overall performance and efficiency. Further, evidence indicates that once companies have achieved a certain scale, then diminishing profitability and productivity would stimulate them to search for new investment opportunities (Bakke and Gu 2017). Therefore, a diversification strategy is more likely to be adopted by larger companies with diminishing value. However, using a sample of Ukraine banks, Mertens and Urga (2001) find that small commercial banks have more efficient performance in terms of cost compared with large-sized banks, and thus detect diseconomies of scale in the banking sector. Similarly, in their investigation of the US bank holding companies, Akhigbe and Stevenson (2010) find that the benefits from scale economies are not sufficient to improve banks’ efficiency, but generate an efficiency discount. Research that examines the effect of income diversification on Chinese banks is just emerging. Zhang (2003) investigates Chinese listed commercial banks and finds that banks’ profitability is highly correlated with diversification levels, with a positive sign. Meanwhile, banks’ operational risk has a non-significant but negative correlation with diversification. Xia and Huang (2017) further suggest that the riskreduction effect is mainly generated from portfolio diversification. Chi (2006) apply DEA to balance sheet data from 14 Chinese banks and find that more diversified banks with higher levels of non-interest activities could gain

5.2 Literature Review

125

large improvements in their efficiency scores. Using the same efficiency estimation approach as Chi et al. (2006), Chen and Chen (2015) expand the sample to city banks in China and find a diversification premium, driven mainly by the accumulation of professionals and technological spillover effects from other business lines. Wei and Liu (2007) introduce the entropy index to evaluate the diversification level in the Chinese banking sector and find a very small positive diversification effect on banks’ efficiency. Xia and Huang (2017) detect a less-significant efficiency premium and present evidence that the increased management and operational costs greatly decrease the efficiency improvement obtained from a cross-selling strategy. Similar results are reported in Liu and Ji’s (2014) study of 45 Chinese banks during the period from 2008 to 2012. By employing the DEA approach, Liu and Ji (2014) find that the expansion of non-interest activities causes devaluation of banks’ efficiency, driven mainly by risk enhancement and increased management cost.

5.3 Variables, Data and Methodology 5.3.1 Variables Measures of Banks’ Efficiency. Two efficiency scores are used in this chapter, namely scores for cost efficiency and for profit efficiency. Both are calculated through stochastic frontier analysis (SFA) deploying the within maximum likelihood estimation (WMLE) method. Each efficiency score ranges from 0 to 1, indicating least to highest efficiency. A commercial bank with an efficiency score of 0.7 is 70% as efficient as the best-performing banks in the sample year. The banking literature includes two main perspectives on the role of commercial banks and the components of inputs and outputs used to estimate efficiency score. The production approach suggests that production units use physical inputs such as capital and labour to supply service to customers to achieve outputs such as taking customer deposits and issuing loans. On the other hand, the intermediation approach treats commercial banks as intermediaries, whose function is to gather funds from the public and transfer these into profitable assets and projects. Owing to issues regarding data availability, this study follows Dong et al. (2016) in choosing the intermediation approach to estimate the efficiency level. This is because the information required under the production approach, such as the number of accounts held by the bank, is not publicly available, whereas the intermediation approach requires accounting-based information that can be found in public annual banking reports. Therefore, this book uses two inputs (xit ) prices, namely the price of total physical capital (TC), which is measured by the ratio of other operating expenses to the book value of fixed assets; and the price of total borrowed funds (TF), which is measured by the ratio of total interest expenses on borrowed funds to total borrowed funds. The outputs (yit ) can be broken down into total loans (TL), other earning assets (OEA), and non-interest income (LA). The total cost used in the model includes both interest

126

5 The Impact of Income Diversification on Chinese Banks …

and operating expenses, including interest expenses, employee benefits, employee salaries and other operating costs. To solve the omitted variables problem of the sample of banks, this study introduces three control variables. Following Dong et al. (2014), we use the total equity capital (z) of the specific banks as a quasi-fixed input in the banking cost function in order to control for banks’ insolvency risk and different risk preferences. In addition, time trend (T ) is used to account for the effects of technical progress, such as the learning-by-doing effect and technical spillover, over time. Measures of Income Diversification. Following Amidu and Wolfe (2013), Gurbuz et al. (2013) and Meslier et al. (2014), we adopt the Herfindahl–Hirschman index (HHI) as the indicator of diversification. It is commonly used in similar research and is calculated as:   DIV = 1 − (INT/TOR)2 + (FEE/TOR)2 + (TRA/TOR)2 + (OTH/TOR)2 (5.1) where INT is the gross interest revenue, TOR is the total operating revenue, COM refers to the ratio of net fee and commission income to total operating income, TRA is the ratio of net trading income to total operating income, and OTH indicates the ratio of net other operating income to total operating income. The investigation of diversification activity considers three types of income, namely, fee and commission income, trading income and other income. Following Köhler (2014), we use the corresponding indexes as proxies. These are the ratio of net fee and commission income to total operating income (COM), ratio of net trading income to total operating income (TRA), and ratio of net other operating income to total operating income (OTH). Measures of Competitiveness. The Lerner index is adopted as the indicator of the level of competitiveness of the banking sector. The Lerner index is defined as the difference between a bank’s price and the marginal cost divided by the price. The price is estimated by the average price of bank production as the ratio of total revenue to total assets (Tan et al. 2017), that is: Lerner = ( p − MC)/ p

(5.2)

The higher the index is, the more market power and competitiveness the bank in question possesses. The marginal cost is the key input for estimating the Lerner index, and it can be calculated by taking the first derivative of the dependent variables in the translog equation. Specifically, following Tan et al. (2017), the marginal cost is estimated on the basis of a translog cost function with signal output (total assets). Because of the data restriction of the labour process, we select two input prices, namely price of capital and price of funds. Also, we use a fixed net-put (equity) and technical changes (using a time trend as a proxy). Other Variables. The other variables include the following:

5.3 Variables, Data and Methodology

127

NIM: This is the net interest margin, indicating the net interest revenue over total earning assets. It is intended to describe the interest-based activities (Lepetit et al. 2008; Busch and Kick 2009; Köhler 2014). ETA: To adjust for banks’ attitude toward efficiency, we adopt the ratio of equity to assets, which describes the degree of total financial leverage and capital adequacy (Stiroh 2004; Pennathur et al. 2012; Gurbuz et al. 2013). CIR: The cost-income ratio is estimated through the operating expenses relative to gross income, which measures banks’ cost structure (Busch and Kick 2009).

5.3.2 Data Sample Our sample comprises an unbalanced panel of 40 Chinese commercial banks from 2005 to 2016, with annual data drawn mainly from BankScope and banks’ annual reports. The sample accounts for 79% of total assets of the Chinese banking industry. Drawing from Basel III and the China Banking Regulatory Commission (CBRC), the banks are divided into three groups: global systemically important banks (GSIBs), domestic systemically important banks (D-SIBs), and other banks that are not classified by the authorities as systemically important (N-SIBs).

5.3.3 Investigation Strategy and Research Methodology 5.3.3.1

Estimation of the Efficiency Scores

We adopt a two-stage strategy to investigate the relationship between bank efficiency and bank diversification in China. First, we estimate the efficiency of the banks. This is achieved by using stochastic frontier analysis (SFA) to evaluate banks’ efficiency scores. Employing a set of statistical techniques for economic modelling of firm behaviour, the SFA explicitly recognizes the existence of firm inefficiency. Its theoretical underpinning can be traced back to Hicks (1935), who claimed that in addition to seeking profit maximization, monopolists may have other motivations that lead to sub-optimality of production. This argumentation has opened a path for research on producers who behave in a less than optimal manner when seeking profit maximization or cost minimization. Aigner et al. (1977) and Meeusen and Van den Broeck (1977) were among the first to apply the theory to empirical estimation of producers’ conduct in the presence of firm inefficiency. The empirical research was initially focused on the production function, and then expanded to the cost function. Subsequently, the research has extended from economics to financial studies (Berger and Humphrey 1997; Kumbhakar and Lovell 2003; Cavallo and Rossi 2002; Kraft et al. 2006; Fenn et al. 2008; Kao and Liu 2009; Feng and Zhang 2012; Dong et al. 2014, 2016).

128

5 The Impact of Income Diversification on Chinese Banks …

For illustration, we start with the production frontier model, which is the empirical departure point for SFA. The production frontier model in log form can be presented as: yit = αi + β  xit + εit ,

(5.3)

εit = νit − u it . where i = 1, …, N indexes firms and t = 1, …, T indexes time periods. In Eq (5.3), yit is the observed scalar output of the producer i at time t; xit represents the vector of N production inputs used by producer i (e.g. labour and capital); β is a vector of technology parameters to be estimated, so that f(x it , β) is known as the production frontier since it indicates the frontier of maximal output for a given set of inputs x i . In addition, αi captures the unobserved heterogeneity in terms of time-invariant effects (of incidental parameters), and εit indicates the error term. This error term is a compound one consisting of two components: νit and − u it . Here, {νit } is the noise component capturing the effects of random shocks affecting the production process. This component introduces stochasticity into the model. The {−uit } component contains non-negative errors representing unobserved inefficiency, which is the salient feature of SFA. To elaborate, let TEi denote the i-th firm’s technical efficiency, measured by the ratio of observed output to maximum feasible output. TEi = 1 means firm i obtains the maximum feasible output, while TEi < 1 indicates that the firm achieves less than its maximum feasible output. So, we have TEi ≤ 1. Further, if we let TEi = −ui , then in exponential form, we have exp TEi = exp {−ui } where ui ≥ 0, given TEi ≤ 1. Plugging exp {−ui } into Eq. 5.1, and recall that it is in log form, we have: yit = αi + β  xit + vit − u it

(5.4)

This sheds further light on Eq. 5.3, showing that Eq. 5.4 is actually an errorcomponent model. Whereas we employ the production frontier model for illustration of the modelling setup and methodology, SFA also examines cost efficiency (Kumbhakar and Lovell 2003) and has been applied to other areas of economics and banking analysis. For functional forms, in addition to the common use of natural logarithms, other forms such as translog functions are also modelled. Depending on the modelling specification, appropriate elements are selected for the cost or profit frontier of x i . The (ui ) component of the composed error can also be production, revenue, profit, or cost inefficiency. Given that the research interest of this chapter is bank efficiency in China, we specify stochastic frontier analysis in terms of both profit and cost efficiency (See Eqs. 5.28–5.31). The underlying model is similar to that of the production frontier model. Some revisions are made so that the examination addresses both the profit frontier and the cost frontier function. In these models, it is the banks rather than

5.3 Variables, Data and Methodology

129

corporate producers that are the profit-maximizers or cost-minimizers. The outputs are the observed total cost or profits of the bank i at time t. Estimation of SFA may be conducted via Greene’s (2005) true fixed-effects approach. However, that approach suffers from the incidental parameters problem, whereby the variance parameters are more likely to be affected under the shortpanel condition (Greene 2005). Belotti and Ilardi (2012) suggest that this may be improved if the panel length is sufficiently large, 15 or greater. However, since the number of time periods in the data sample is only 12, Greene’s model is not suitable for the estimation, due to the limited data. Instead, we adopt the method of within maximum likelihood estimation (WMLE) introduced by Chen et al. (2014) based on fixed-effects estimation. More specifically, Chen et al.’s (2014) estimation is based on the within-transformed model using the maximum likelihood method. This procedure does not suffer from the ‘incidental parameters’ problem because within-transformation removes the incidental parameters and the firm effects are fixed, such that: z˜ it = z it − z¯ i

(5.5)

where for each panel i and any variable (z), the individual mean (¯z i ) is subtracted  from the observed value in period t (z it ) which can be defined as z¯ i = T1 t z it . Therefore, deviations from the means (˜z it ) can be used in the model. The resulting formulation is free of αi ; specifically, α˜ i = 0. Thus, the fixed-effects stochastic frontier model with within transformation is of the form: y˜it = β  x˜it + ε˜ it ,

(5.6)

ε˜ it = ν˜ it − u˜ it ,

(5.7)

  νit ∼ I I DN 0, σv2 ,

(5.8)

  u it ∼ I I DFu σu2 , i = 1, . . . , n, t = 1 . . . , T,

(5.9)

where error term εit indicates the difference between the idiosyncratic error term νit and inefficiency component u it . νit and u it are independently distributed. The inefficiency term u it is distributed according to Fu with a specific non-normal distribution, and we assume that it is half-normal, whereas νit is normally distributed. Let λ = σσuv and σ 2 = σu2 + σv2 . Then, the density of the composed error should be: f(ε) =



λε 2 ε ϕ φ − σ σ σ

(5.10)

The distribution of Eq. 5.10 is a member of the skewed normal family introduced by Azzalini (1985), which suggests that the Closed Skew Normal (CSN) distribution

130

5 The Impact of Income Diversification on Chinese Banks …

is suitable in the stochastic frontier context. The distribution of the composed error can be written as:

λ (5.11)

it ∼ C S N1,1 0, σ 2 , − , 0, 1 σ The density of C S N p,q distribution includes a p-dimensional pdf and a qdimensional cdf of a normal distribution.  With panel data, the distribution of T-dimensional vector i = ( i1 , . . . , i T ) can be rewritten as:

λ

i ∼ C S N T,T 0T , σ 2 IT , − IT , 0T , IT σ

(5.12)

where I  is the identity matrix, in which the vector includes the mean of errors, i.e.

¯i = T1 t εit and ˜i∗ = (˜ i1 , . . . , ˜i,T −1 ) , which indicates the vector of the first   T − 1 deviations from the mean ˜i∗ . The likelihood function is parameterized in terms of β, λ = σu /σv and σ 2 = σu2 + σv2 , where the incidental parameters problem is avoided and inefficiency is allowed to be time-varying. By adopting the point estimator of Battese and Coelli (1988), the composed error can be written as:

it = yit − yˆit = yit − βˆ  xit − αˆ i

(5.13)

In Eq. 5.13, the value of αˆ i can be estimated through the method proposed by Chen et al. (2014), where  αˆ iM

ˆ

= y¯i − β x¯i +

2 σˆ u π

(5.14)

where βˆ  and σˆ u are the within maximum likelihood estimation (WMLE) estimates. Then, based on Battese and Coelli (1988), the efficiency term can be calculated as: EF F it = E(exp(−u it )| it )

5.3.3.2

(5.15)

Dynamic Doubly Censored Tobit Model

In the second step, we employ the limited dependent variable model to investigate the effects of diversification on bank efficiency, since each set of the efficiency scores is limited to values between 0 and 1. Furthermore, in the model used here, the distribution of the dependent variable is expected to be half normal rather than normal, and the error terms cannot meet the assumption of a normal distribution. Thus, noncensored estimates such as OLS will be biased and inappropriate for estimation, since in OLS, the dependent variable can take on a negative or positive real value. The

5.3 Variables, Data and Methodology

131

consequences of ignoring fractionality may result in biased estimation and inconsistent parameter estimates (Greene 1980). Therefore, we set up the dynamic doubly censored Tobit model with a left censored bound of zero and a right censored bound of one to regress bank-level efficiency scores against banks’ income diversification and several control variables. The dynamic Tobit estimation was developed by Elsas and Florysiak (2015), based on Loudermilk (2007). With this estimator, the distribution of the unobserved fixed effects is assumed to be conditional on the initial value of the dependent variable and the time averages of the exogenous explanatory variables (Wooldridge 2005). In essence, this modelling strategy allows for firm heterogeneity. An additional important feature of this approach is that, unlike the dynamic Tobit model of Loudermilk (2007), the approach by Elsas and Florysiak (2015) does not require balanced panel data and is robust to missing data in unbalanced panels. The Elsas and Florysiak (2015) dynamic Tobit model is suitable for unbalanced dynamic panel data with a fractional dependent variable (DPF estimator) and can capture fixed effects in estimating the unobserved, time-invariant firm heterogeneity. In this approach, the DPF estimator is a doubly censored Tobit estimator employing a latent variable specification to estimate the fractional nature of the dependent variable. The specification includes corner observations at 0 and 1, with a lagged dependent variable. In its general form, we have:   yit∗ = z it γ + g yi,t−1 ρ + ci + u it ,

(5.16)

    u it | z it , yi,t−1 , . . . , yi0 , ci ∼ N 0, σu2 .

(5.17)

The observable doubly censored dependent variable with two possible corner outcomes is as follows: ⎧ ⎨ 0 i f yit∗ ≤ 0 (5.18) yit = yit∗ i f 0 < yit∗ < 1 ⎩ 1 i f yit∗ ≥ 1 where z it refers to strictly exogenous regressors, ci indicates the unobserved effect, and u it is a normally distributed error term; yit∗ is the unobserved latent variable, which is set equal to zero when it is below zero and to one when it is greater than one. The joint density of (yi1 , . . . , yi T ) given (yi0 , z i , ci ) is given by: f (yi1 , . . . , yi T |yi0 , z i , ci ) =

T 

f t (yit |wit , ci ; θ )

(5.19)

t=1

As the density of yit = (yi1 , . . . , yi T ) given (yi0 , z i , ci ), to proceed with the estimation it is necessary to specify the density of ci given (yi0 , z i , ci ). Elsas and

132

5 The Impact of Income Diversification on Chinese Banks …

Florysiak’s DPF estimator specifies a conditional distribution for unobserved heterogeneity ci based on Loudermilk (2007). The unobserved fixed-effects distribution is assumed to be: ci = α0 + α1 yi0 + z¯ it α2 + εi

(5.20)

 is the time-series average of z it . where the error term εi is normally distributed, and z¯ i,t Unlike the Tobit estimation developed by Loudermilk (2007), rather than including the term z it , the DPF estimator assumes that the fixed effects distribution depends on time-series averages of the exogenous variables; hence, it does not require the fixed effects to depend on a balanced panel. The substitution for ci is produced by:

  P yit = 0|wit , z¯ it , yi0 , αi

 −¯z it γ − gi,t−1 ρ − α0 − α1 yi0 − α2 z¯ i − αi = σu   P yit = 1|wit , z¯ it , yi0 , αi

 z¯ it γ + gi,t−1 ρ + α0 + α1 yi0 + α2 z¯ i + αi − 1 = σu

(5.21)

(5.22)

and ∂ P(yit ≤ y|wit , z¯ it , yi0 , αi ) ∂y

yit − z¯ it γ − gi,t−1 ρ − α0 − α1 yi0 − α2 z¯ i − αi 1 = φ σu σu

(5.23)

Therefore, the log-likelihood function can be estimated by integrating the density of (yi1 , . . . , yi T ) given (yi0 , z i , ci ) against the distribution of αi : L=

n  i=1

log

  T 



f t yit |wit , z¯ it , yi0 , αi ; θ

t=1

   1 a da φ σa σa

(5.24)

After iterated expectations, defining    1 =  −wit β − α0 − α1 yi0 − α2 z¯ i /σu ,    2 =  1 − wit β − α0 − α1 yi0 − α2 z¯ i /σu ,    φˆ 1 = φ −wit β − α0 − α1 yi0 − α2 z¯ i /σu ,    φˆ 2 = φ 1 − wit β − α0 − α1 yi0 − α2 z¯ i /σu and σv = σu + σa . 



The conditional mean function can be described as:

(5.25)

5.3 Variables, Data and Methodology

133

  wit β + α0 + α1 yi0 + α2 z¯ i − 1 r wit , z¯ it , yi0 ; θ = Φ σv      + wit β + α0 + α1 yi0 + α2 z¯ i Φ 2 − Φ 1 + σv Φ 1 − Φ 2 , 







(5.26) and estimation of average partial effects is given by:   ∂r wit , z¯ i , yi0 ; θ |θ=θˆ ∂w j      N   1 − σv 1  1 − σv  φˆ 2 − wit β + α0 + α1 yi0 + α2 z¯ i φˆ 2 − φˆ 1 = βj 1 + N i=1 σv σv (5.27) As the distribution of ci is specified in terms of observables and a normally distributed error term, partial effects on P(yit = 0|wit , ci ) and P(yit = 1|wit , ci ) can then be computed.

5.4 Empirical Estimation 5.4.1 Estimating Efficiency Scores for Chinese Banks 5.4.1.1

Specification for the Empirical Model of Stochastic Frontier Analysis

We first establish the specification for SFA, and considering the relatively short panel length of the data for Chinese banks, we apply the analysis with the method of within maximum likelihood estimation (WMLE). In the estimation, the cost and profit frontier models are expressed in Eqs. 5.28 and 5.29 (our empirical cost and profit frontier models are shown in Eqs. 5.30 and 5.31): T C it = f (Yit , Wit , Z it ; β) + νit + u it i = 1, . . . I, t = 1, . . . T

(5.28)

T P it = f (Yit , Wit , Z it ; β) + νit − u it i = 1, . . . I, t = 1, . . . T

(5.29)

where following Lensink and Meesters (2014), the functional form of f (Yit , Wit , Z it ; β) is estimated by translog form. T Cit and T Pit refer to the observed total cost and profits before tax for bank i at time t; Yit and Wit represent the vectors of output and input prices for a specific bank; Z it refers to a vector of control variables, and β is a vector of technology parameters. In SFA, the error term can be disentangled into two elements: νit is the measurement error and random effects,

134

5 The Impact of Income Diversification on Chinese Banks …

which are assumed to follow a normal distribution, i.e. vit ∼ iid N (0, σv2 ); and u it is the inefficiency term, which is assumed to follow a half-normal distribution, i.e. u it ∼ N + (0, σv2 ). wit is the effect of unobserved factors, which follows a truncated normal distribution with zero mean and constant variance. Because it is necessary to ensure that all dependent variables are positive, we follow Dong et al. (2016) and delete all observations of the profit variable with a negative sign. SFA uses a parametric approach, which requires specification of the functional form of the production function and the distribution of its error terms. According to the duality theorem, the cost function must be linearly homogeneous in input prices, whereas continuity requires that the second-order parameters must be symmetric. Hence, we scale the total costs and input price by one price, w jit , to impose a linear homogeneity restriction on the model (Dong et al. 2016). In addition, there are standard symmetry restrictions, where γ jk = γk j and βnm = βmn . Thus we specify the cost function as: ln

T Cit wkit

= αi +

3 

βm lnymit +

m=1

3 3 1   βmn lnymit lnynit 2 m=1 n=1

3 1 1  + γ1 ln(w jit /wkit ) + γ2 ln(w jit /wkit )2 + ψm j lnymit (w jit /wkit ) 2 2 m=1

+ φ1 ln Z it +

+

1 2 + φ2 ln Z it 2

3 

ln ymit ln Z it + ξln(w jit /wkit )ln Z it + θ1 T

m=1

3  1 φ2 T 2 + km lnymit T + ρln(w jit /wkit )T + ηln Z it T + νit + u it 2

(5.30)

m=1

As we utilise fixed-effect estimation, in this equation, αi is the unobserved “heterogeneity” of utility i, which is treated as fixed; the dependent variable of the cost function ln(T Cit ) refers to logarithm of total cost, including labour, interest, and other costs; lnymit indicates the logarithm of the output of a specific bank; lnw jit indicates the logarithm of input price of a specific bank; ln E it refers to the environmental variable, which is the logarithm of total equity of a specific bank; and u it is the inefficiency term, with an explicit function of environmental variables that impact each bank’s best performance. With regards to profit efficiency, we utilise an alternative measure, which is calculated using a translog functional model similar to that used for the cost efficiency. Instead of total cost, we use the logarithm of profit before tax ln(T Pit ), along with the same independent variables as used in the cost function. Hence, we specify the profit function as:

5.4 Empirical Estimation

ln(T Pit ) = αi +

3 

135

βm lnymit

m=1

+

3 2 3  1  βmn lnymit lnynit + γ j lnw jit 2 m=1 n=1 j=1

+

2 3 2 2 1  1  γ jk lnw jit lnwkit + ψm j lnymit lnw jit + φ1ln Z it 2 j=1 k=1 2 m=1 j=1

  1 + φ2 ln Z it2 + ln ymit ln Z it + ξ J lnw jit ln Z it + θ1 T 2 m=1 j=1 3

2

3 2   1 km ln ymit T + ρ J lnw jit T + ηln Z it T + νit − u it + φ2 T 2 + 2 m=1 j=1

(5.31)

5.4.1.2

Empirical Results of SFA

Table 5.1 presents the estimation results of the cost and profit frontier models using maximum likelihood techniques. As my main interest here is to estimate the diversification effects, we do not discuss the estimated coefficients on other variables of the frontiers in detail. However, stochastic frontier analysis fulfils the theoretical requirements for a valid cost function. More specifically, the tests in terms of the monotonicity of the cost function are satisfied, as the estimates for ∂ ln(TC)/∂ ln(Q i ) and ∂ ln(TC)/∂ ln(Wi ) are all positive, thus indicating that the cost function is non-decreasing in outputs and input prices. The yearly mean efficiency estimations from 2005 to 2016 for the full sample and three sub-groups, namely, G-SIBs, D-SIBs and N-SIBs, are plotted in Figs. 5.1 and 5.2. It can be observed that the average of both cost and profit efficiency scores exhibit an increasing tendency from 2005 to 2008 and then begin to decline, reaching the lowest point in 2010. Subsequently, the efficiency of Chinese commercial banks improves steadily, to achieve a relatively high point in 2015. The three sub-groups each follow a similar trend to that of the overall banking sector, exhibiting a general increase from 2005 to 2008 and then a decrease, reaching the lowest point in 2010, and then starting to rise once again. In more detail, in 2005, D-SIBs were the most efficient bank group in the Chinese banking market; however, their efficiency decreased dramatically after 2006. G-SIBs generally maintained a cost efficiency score midway between those of the other two groups. It seems that the efficiency of those banks does not benefit greatly from their scale of assets or scope of business.

136

5 The Impact of Income Diversification on Chinese Banks …

Table 5.1 Parameter estimates of the cost and profit frontiers Cost frontier

Profit frontier

Variables

Coefficient Standard error Variables

Coefficient Standard error

lny1

0.486**

0.201

lny1

0.178

0.203

lny2

0.390*

0.230

lny2

−0.435*

0.231

lny3

0.449**

0.212

lny3

0.899***

0.185

0.5 * (lny1)ˆ2

0.157***

0.021

0.5 * (lny1)ˆ2

0.192***

0.020

lny1 * lny2

−0.333*** 0.018

lny1 * lny2

−0.271*** 0.009

lny1 * lny3

0.114**

0.048

lny1 * lny3

0.021*

0.011

0.5 * (lny2)ˆ2

0.24***

0.023

0.5*(lny2)ˆ2

0.385***

0.017

lny2 * lny3

−0.005

0.023

lny2 * lny3

−0.075*** 0.011

0.5 * (lny3)ˆ2

−0.041

0.032

0.5 * (lny3)ˆ2

−0.004

0.016

ln(w1/w2)

1.083***

0.155

lnw1

1.392***

0.398

lnw2

−0.599*** 0.172

0.5 * (ln(w1/w2))ˆ2 0.039**

0.017

0.5 * (lnw1)ˆ2 0.068

0.058

−0.076**

0.031

0.5 * (lnw2)ˆ2 0.051***

0.019

lnw1 * lnw2 lny1 * ln(w1/w2)

−0.070*** 0.021

lny1 * lnw1 lny1 * lnw2

0.069**

0.027

lny2 * ln(w1/w2)

−0.004

lny2 * lnw1

−0.114**

0.051

lny2 * lnw2

−0.009

0.023

lny3 * lnw1

0.091**

0.037

lny3 * lnw2

0.021

0.018

lny3 * ln(w1/w2)

0.037*

0.022 0.019

−0.142*** 0.044

lnz

−0.896*** 0.168

lnz

−0.176

0.316

0.5 * (lnz)ˆ2

−0.037

0.023

0.5 * (lnz)ˆ2

−0.032

0.032

lnz * lny1

0.072***

0.026

lnz * lny1

0.064***

0.020

lnz * lny2

0.106***

0.025

lnz * lny2

−0.030**

0.012

lnz * lny3

−0.096*** 0.014

lnz * lny3

0.038***

0.010

lnz * ln(w1/w2)

0.039*

lnz * lnw1

0.159***

0.060

lnz * lnw2

−0.072**

0.031

0.023

T

0.283***

0.036

T

0.219***

0.070

0.5 * (Tˆ2)

0.001

0.002

0.5 * (Tˆ2)

0.003

0.003

T * ln(w1/w2)

0.005

0.004

T * lnw1

0.006

0.008

T * lnw2

0.008*

0.004

T * lny1

−0.024*** 0.004

T * lny1

−0.020

0.015

T * lny2

−0.021*** 0.006

T * lny2

0.001

0.011

T * lny3

0.005

0.005

T * lny3

−0.005

0.005

T * lnz

0.029***

0.007

T * lnz

0.016

0.021

Constant

2.536***

0.658

Constant

1.788***

0.598

COST EFFICIENCY

5.4 Empirical Estimation

137

0.95 0.9 0.85 0.8 0.75 0.7 2005

2006

2007

2008

G-SIBs

2009

D-SIBs

2010

2011

N-SIBs

2012

2013

2014

2015

2016

2015

2016

Whole Sample

PROFIT EFFICIENCY

Fig. 5.1 Yearly mean cost efficiency for the whole sample and three sub-groups

0.96 0.94 0.92 0.9 0.88 0.86 0.84 0.82 0.8 2005

2006

2007

2008

G-SIBs

2009 D-SIBs

2010

2011 N-SIBs

2012

2013

2014

Whole

Fig. 5.2 Yearly mean profit efficiency for the whole sample and three sub-groups

5.4.2 Determination of Diversification Effects on Cost and Profit Efficiency We apply an estimator designed to be unbiased in the context of unbalanced dynamic panel data, with a fractional dependent variable to regress the efficiency scores on the diversification level. This is within the family of censored regression models. The particular functional form we adopt is the dynamic Tobit model, as against the Probit model. The dynamic Tobit model is extensively discussed in Hu (2002), Wooldridge (2005) and Li and Zheng (2008). According to Wooldridge (2005), the dynamic Tobit model is described as:     yit = max 0, z it γ + g yi,t−1 ρ + ci + u it    u it  yi,t−1 , . . . yi0 , z i , ci ∼ Normal 0, δ 2 u

(5.32)

where yit is observed response variable of interest on the ith agent in time period t which depends on the explanatory variables z it , the lags of the dependent variable yi,t−1 and the unobserved individual heterogeneity ci . u it is the error terms, which are  T , ci . assumed to be i.i.d. normally distributed conditional on yi,t−1 . . . yi,0 , {z it }t=2

138

5 The Impact of Income Diversification on Chinese Banks …

In a dynamic panel Tobit model, researchers often take the following form (Hu 2002): yit∗ = xit β + yi,t−1 λ + it   yit = max yit∗ , 0

it = ai + u it , i = 1, . . . N ; t = 1, . . . T

(5.33)

where yit∗ is latent dependent variable, yi,t−1 is first lag of the observed dependent variable, xit is a vector of exogenous variables, β are the regression coefficients, λ is the coefficient on the lagged dependent variable. Further, the component ai is unobserved individual effect and u it is the error terms. In the panel data Tobit model, ∗ λ introduces the dynamics into the system. For our research the variable of yi,t−1 interests, we have: EFFit = λEFFit−1 + (DIVit , Lernerit , NIMit , ETAit , CIRit ) β + αi + u it (5.34) Empirical results which are estimated by Eq. 5.34 are reported in Tables 5.2 and 5.3 for whole sample and in Tables 5.4, 5.5, 5.6, 5.7, 5.8 and 5.9 for three Chinese banking groups (G-SIBs, D-SIBs and N-SIBs). In this dynamic panel Tobit model, the dependant variables are EF F it which is represented by two different measurements of banks’ efficiency, namely, cost and profit efficiency. In addition to the lagged dependent variables in the model, the vector of exogenous variables contains DIVit , Lernerit , NIMit , ETAit , CIRit ; where DIVit is the variable of our main interest, which captures the level of diversification, represented respectively by the Herfindahl-Hirschman Index and the shares of three non-interest components over total income; Lernerit is the Lerner index; NIMit is the ratio of net interest revenue to total earning assets; ETAit is the ratio of equity over total assets; and CIRit is the cost-to-income ratio.

5.4.2.1

Income Diversification and Efficiency: Whole Sample

Having estimated the efficiency scores, we next apply the dynamic Tobit model with the DPF estimator to investigate their determinants. In addition to the above input price and output variables, a number of variables are included to explain the efficiency scores. Table 5.2 presents the results for the Chinese banking industry as a whole. As can be seen from the table, both the cost and profit efficiency exhibit a consistent sign, with negative coefficients at the 1% significance level. This outcome is consistent with Cheng (2015), who finds that income diversification could decrease both cost and profit efficiency in the Chinese banking sector. A plausible explanation for this effect is that the overall level of operational ability in the Chinese banking sector is low and the internal capital market is inefficient. As a result, internal reallocation of resources may have led to over-investment or under-investment, which would increase the costs of coordination and management, leading to inefficiency.

5.4 Empirical Estimation Table 5.2 Income diversification and banks’ efficiency for the whole sample, 2005–2016

139

I_(t − 1) HHI

Cost efficiency

Profit efficiency

0.518***

0.828***

(0.051)

(0.03)

−0.143**

−0.072**

(0.059)

(0.031)

Lerner

−0.276***

−0.001

(0.048)

(0.033)

NIM

2.135***

0.003

(0.698)

(0.451)

0.662***

0.593***

ETA

(0.231)

(0.163)

CIR

−0.041

0.044

(0.051)

(0.032)

Constant

0.460***

0.121***

(0.064)

(0.036)

Log likelihood

579.193

614.549

LR test (p-value)

0.000

0.000

Observations

413

413

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost efficiency and profit efficiency, which are estimated via stochastic frontier analysis. .I(t − 1) refers to the dependent variables lagged by one period. HHI indicates income diversification using the Herfindahl-Hirschman index with the components of interest income and three component activities under non-interest income. Lerner indicates the Lerner index calculated by (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity/total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

One of the control variables employed is the Lerner index, which measures a bank’s level of market power. In the estimation results, both the cost and profit efficiency scores have a negative correlation with the Lerner index, indicating that the higher the market power of the bank is, the less efficient the bank will become. This result is consistent with the notion that banks tend to pursue a ‘quiet life’; that is, banks with higher monopoly power seem to allow costs to rise as a consequence of slack management. When market power prevails, managers may pursue objectives other than profit maximization, and they do not have incentives to work hard to keep costs under control, a situation that leads to a reduction in cost efficiency (Koetter et al. 2008; Delis and Tsionas 2009; Ariss 2010).

140

5 The Impact of Income Diversification on Chinese Banks …

Table 5.3 Results for three components of non-interest income and bank efficiency for Chinese banks, 2005–2016

I_(t − 1) COM

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

0.484***

0.822***

0.522***

0.823***

0.516

0.827***

(0.043)

(0.031)

(0.059)

(0.029)

(0.404)

(0.028)

−0.454***

−0.152***

(0.083)

(0.054) 0.471**

0.136

(0.185)

(0.114) −2.234**

−0.091

TRAD Other

(1.078)

(0.196)

−0.288***

−0.003

−0.302***

−0.024

−0.327

−0.02

(0.046)

(0.032)

(0.047)

(0.03)

(0.435)

(0.03)

1.982***

0.075

3.323***

0.539

2.352

0.423

(0.621)

(0.436)

(0.672)

(0.408)

(3.105)

(0.417)

ETA

0.781***

0.594***

0.551**

0.308**

0.307

0.358

(0.216)

(0.161)

(0.229)

(0.145)

(1.184)

(0.142)

CIR

−0.076

0.04

−0.011

0.037

−0.055

0.039

(0.049)

(0.032)

(0.049)

(0.03)

(0.344)

(0.03)

0.506***

0.125***

0.401***

0.123***

0.397

0.120***

(0.056)

(0.036)

(0.063)

(0.033)

(0.688)

(0.033)

Log likelihood

586.916

615.78

564.058

706.621

405.880

719.503

LR test (p-value)

0.000

0.000

0.000

0.000

0.000

0.000

Observations

413

413

404

404

411

411

Lerner NIM

Constant

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost efficiency and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. COM is the ratio of net fee and commission incomes to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. Lerner is the Lerner index calculated as (bank price − marginal cost)/bank price, NIM is the net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

It is conceivable that components of the non-interest business may perform differently than the overall non-interest activities. To find further evidence for diversification effects across different components of non-interest income, we divide the non-interest income into three categories, namely, fee and commissions, trading, and other income; the results are reported in Table 5.3.

5.4 Empirical Estimation Table 5.4 Income diversification and banks’ efficiency for Chinese G-SIBs, 2005–2016

141

I_(t − 1) HHI

Cost efficiency

Profit efficiency

0.984***

0.797***

(0.035)

(0.04)

−0.613***

−0.236***

(0.055)

(0.051)

Lerner

−0.642***

−0.453***

(0.069)

(0.07)

NIM

1.227

−1.730***

(0.802)

(0.675)

4.200***

2.213***

ETA

(0.391)

(0.336)

CIR

−0.001

−0.401***

(0.069)

(0.067)

Constant

0.142**

0.490***

(0.061)

(0.067)

Log likelihood

365.275

236.676

LR test (p-value)

0.000

0.000

Observations

42

42

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost efficiency and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. HHI indicates income diversification using the Herfindahl-Hirschman index with the components of interest income and three component activities under non-interest income. Lerner indicates the Lerner index calculated as (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

As can be seen from Table 5.3, all lagged dependent variables exert a significant and positive effect on the current efficiency level. In particular, the results indicate that Chinese banks could reap efficiency advantages from a shift towards trading activities. However, both the fee-based and other activities could bring an efficiency discount in the process of income diversification.

5.4.2.2

Income Diversification and Efficiency Across Banking Groups

Several studies claim that the strength of the relationship between income diversification and banks’ efficiency could be greatly affected by bank business scale. Specifically, large banks should have a higher share of non-interest income and a better

142 Table 5.5 Income diversification and banks’ efficiency for Chinese D-SIBs from 2005 to 2015

5 The Impact of Income Diversification on Chinese Banks …

I_(t − 1) HHI

Cost efficiency

Profit efficiency

0.842***

1.003***

(0.025)

(0.016)

−0.162***

−0.105***

(0.056)

(0.045)

Lerner

−0.278***

−0.086***

(0.037)

(0.029)

NIM

1.705**

−2.938***

(0.837)

(0.663)

0.899***

1.179***

ETA

(0.3)

(0.247)

CIR

−0.045

−0.01

(0.06)

(0.05)

Constant

0.199***

0.074**

(0.046)

(0.033)

Log likelihood

380.025

211.442

LR test (p-value)

0.000

0.000

Observations

97

97

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost efficiency and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. HHI indicates income diversification using the Herfindahl-Hirschman index with the components of interest income and three component activities under non-interest income. Lerner indicates the Lerner index calculated as (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

cost management capacity, whereas at the same time, given the agency problem, managing organized chaos would lead to a reduction in operational efficiency. To investigate whether the diversification effect on banks’ efficiency varies with bank size, we test the effects across the three groups: G-SIBs, D-SIBs and N-SIBs.1

1 As

the small sample size might result in an incidental parameters problem, this book also applies a robustness test by using dummy variables to catalogue the three sub-groups. The robustness test results are reported in the Appendix.

5.4 Empirical Estimation Table 5.6 Income diversification and banks’ efficiency for Chinese N-SIBs, 2005–2016

143

I_(t − 1) HHI

Cost efficiency

Profit efficiency

0.523***

0.454*

(0.100)

(0.241)

0.019

0.009

(0.102)

(0.157)

Lerner

−0.299***

−0.045

(0.104)

(0.16)

NIM

4.507***

5.657**

(1.337)

(2.633)

0.123

0.01

ETA

(0.459)

(0.835)

CIR

−0.044

0.042

(0.096)

(0.126)

Constant

0.437***

0.394

(0.115)

(0.251)

Log likelihood

49.290

1365.163

LR test (p-value)

0.000

0.000

Observations

274

274

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost efficiency and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. HHI indicates income diversification using the Herfindahl-Hirschman index with the components of interest income and three component activities under non-interest income. Lerner indicates the Lerner index calculated as (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

China’s Global Systemically Important Banks Table 5.4 reports the results of estimating the dynamic Tobit model with the DPF estimator with particular reference to G-SIBs. The findings show that banks’ income diversification has significant and negative effects on both cost and profit efficiency of G-SIBs, thus providing evidence that diversified banks incur an efficiency discount compared with banks that focus on traditional sources of interest income. The coefficient on cost efficiency is −0.613, whereas that on profit efficiency is −0.236, both significant at the 1% level. This result indicates that higher income diversification would lead to greater discounts to cost efficiency than to profit efficiency.

144

5 The Impact of Income Diversification on Chinese Banks …

Table 5.7 Results for three components of non-interest income and bank efficiency for G-SIBs from 2005 to 2016

I_(t − 1) COM

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

0.811***

0.643***

1.099***

1.051***

0.927***

0.758***

(0.021)

(0.032)

(0.068)

(0.052)

(0.029)

(0.03)

1.663**

1.341***

(0.738)

(0.52)

−1.321*** −1.085*** (0.072)

(0.113)

TRAD

−1.193*** −1.415***

Other

(0.306) Lerner

(0.19)

−0.701*** −0.274*** −1.155*** −1.125*** −0.736*** −0.302*** (0.046)

(0.043)

0.391 (0.498)

ETA

(0.248)

(0.432)

(0.282)

CIR

−0.757*** −0.649*** −0.289*

0.328***

0.009

−0.072

(0.05)

(0.057)

(0.169)

(0.056)

(0.085)

(0.05)

0.753***

0.795***

0.249*

0.263***

0.12*

0.306***

(0.047)

(0.061)

(0.149)

(0.07)

(0.066)

(0.044)

Log likelihood

859.121

478.811

200.631

164.166

462.949

620.381

LR test

0.000

0.000

0.002

0.000

0.002

0.000

42

41

41

42

42

NIM

Constant

(0.031)

(0.07)

(0.042)

−4.141*** 7.303***

0.844

1.794**

−3.794***

(0.585)

(1.38)

(0.573)

(0.838)

(0.55)

1.825***

1.901***

0.938

0.111

3.511***

3.044***

(0.165)

(0.198)

(0.683)

Observations 42

(0.138)

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. COM is the ratio of net fee and commission income to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. Lerner indicates the Lerner index calculated as (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

In addition, the Lerner index also exhibits a negative effect on banks’ efficiency level, which is consistent with the results for the whole sample. Meanwhile, for this sub-group, ETA maintains a significant positive coefficient correlated with the banks’ efficiency; thus, the capital adequacy in G-SIBs is helpful to improve their efficiency.

5.4 Empirical Estimation

145

Table 5.8 Results for three components of non-interest income and bank efficiency for Chinese D-SIBs from 2005 to 2016

I_(t − 1) COM

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

0.953***

1.001***

0.959***

1.005***

0.964***

1.023***

(0.015)

(0.016)

(0.015)

(0.016)

(0.015)

(0.016)

2.038***

1.272***

(0.464)

(0.381) −0.143

−1.215***

(0.413)

(0.338)

−0.293*** −0.158** (0.084)

(0.072)

TRAD Other

−0.239*** −0.094*** −0.348*** −0.16***

−0.269*** −0.109***

(0.034)

(0.028)

(0.033)

2.171***

−2.861*** 3.046***

−2.355*** 2.703***

(0.769)

(0.657)

(0.778)

(0.65)

(0.772)

(0.64)

ETA

0.846***

1.104***

0.176

0.691***

0.395*

1.039***

(0.252)

(0.231)

(0.221)

(0.19)

(0.229)

(0.195)

CIR

−0.003

−0.012

0.041

0.010

0.040

0.048

(0.057)

(0.05)

(0.056)

(0.048)

(0.059)

(0.05)

0.046

0.076**

0.041

0.076**

0.02

0.042

(0.039)

(0.033)

(0.039)

(0.033)

(0.039)

(0.033)

Log likelihood

662.498

494.027

650.344

490.229

656.521

498.018

LR test

0.000

0.000

0.000

0.000

0.000

0.000

97

95

95

97

97

Lerner NIM

Constant

Observations 97

(0.038)

(0.031)

(0.027) −2.893***

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. COM is the ratio of net fee and commission income to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. Lerner indicates the Lerner index calculated as (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

China’s Domestic Systemically Important Banks We now move to examine the efficiency effect of diversification for China’s domestic systemically important banks (D-SIBs). Table 5.5 reports the results. Next, we examine the regression results for other variables. First, a significantly negative Lerner index indicates that a higher level of monopoly power for a specific bank could be related to lower efficiency. For D-SIBs, we obtain a greater coefficient

146

5 The Impact of Income Diversification on Chinese Banks …

Table 5.9 Results for three components of non-interest income and bank efficiency for Chinese NSIBs from 2005 to 2016

I_(t − 1) COM

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

Cost efficiency

Profit efficiency

0.374***

0.380***

0.375***

0.382***

0.376***

0.413***

(0.053)

(0.057)

(0.049)

(0.058)

(0.051)

(0.064)

−0.321**

−0.214***

(0.138)

(0.078) 0.270*

0.052

(0.153)

(0.121) 0.334

0.028

TRAD Other

(0.31)

(0.132)

−0.230***

−0.009

−0.232***

−0.008

−0.230***

−0.011

(0.061)

(0.037)

(0.05)

(0.039)

(0.055)

(0.043)

2.200***

0.295

2.137***

0.638

2.772***

0.504

(0.675)

(0.442)

(0.581)

(0.455)

(0.662)

(0.508)

ETA

0.477*

0.300*

0.486**

0.219

0.374*

0.325*

(0.247)

(0.156)

(0.206)

(0.161)

(0.225)

(0.184)

CIR

−0.043

0.001

−0.062

0.001

−0.045

−0.012

(0.051)

(0.032)

(0.042)

(0.033)

(0.047)

(0.036)

0.583***

0.562***

0.574***

0.545***

0.556***

0.524***

(0.064)

(0.059)

(0.054)

(0.06)

(0.059)

(0.067)

315.665

486.478

351.333

473.868

386.568

257.604

Lerner NIM

Constant Log likelihood LR test

0.000

0.000

0.000

0.000

0.000

0.000

Observations

274

274

268

268

272

272

This table reports the results of the dynamic Tobit estimation. Our dependent variables are cost and profit efficiency, which are estimated via stochastic frontier analysis. It-1 refers to the dependent variables lagged by one period. COM is the ratio of net fee and commission income to total operating income; TRA is the ratio of net trading income to total operating income; OTH is the ratio of net other operating income to total operating income. Lerner indicates the Lerner index calculated as (bank price − marginal cost)/bank price, NIM indicates net interest revenue over total earning assets, ETA refers to equity over total assets, and CIR indicates the cost-income ratio. Figures in brackets are standard errors. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively

on the Lerner index, indicating that for smaller-sized banks, the negative influence from a higher monopoly power would be smoothed compared with the situation for G-SIBs, such that they experience lower discounts with increased monopoly power. Secondly, the equity-to-assets ratio, which measures the financial leverage, exerts a positive effect on banks’ efficiency.

5.4 Empirical Estimation

147

Other Chinese Banks Table 5.6 reports the results for other Chinese banks. As can be seen from the table, these results differ from those for the other two sub-groups. Here, the findings indicate a significant relationship between the diversification index (HHI) and efficiency scores of Chinese N-SIBs. This suggests that small banks may possess operational advantages that yield higher efficiency, with risk management and project management ensuring efficiency when there is an increase in high-technology requirements and highly leveraged non-interest products (Girardone et al. 2004; Kumbhakar and Wang 2007). As reported in Tables 5.4, 5.5 and 5.6, the results reveal a diversification discount to banks’ efficiency, which eventually becomes a benefit, across the three categories of G-SIBs, D-SIBs and N-SIBs. It is plausible that the differences in results across the three groups are due to bank size, from large G-SIBs to medium-sized D-SIBs and smaller N-SIBs.

5.4.2.3

Effects of Diversification on Efficiency with Components of Non-interest Activities

We also examine the effects of different types of non-interest activities on the efficiency of Chinese banks. The results for G-SIBs are reported in Table 5.7. For trading income, we find positive coefficients, where an increase in scale of trading income would improve banks’ efficiency level. However, commissions and other activities present negative coefficients for both cost and profit efficiency scores at the 1% significance level, indicating that a higher reliance on fee-based and other income is associated with a decrease in banks’ efficiency. Table 5.8 presents the results for D-SIBs. As can be seen from the table, noninterest business components exert different effects on banks’ efficiency scores. Specifically, fee and commissions and other activities have a negative effect on both cost and profit efficiency. The results for all three sub-activities exhibit a similar sign and direction as those for G-SIBs. Table 5.9 presents the effects of three components of non-interest activities on bank efficiency for N-SIBs. By comparing Table 5.9 with Tables 5.7 and 5.8 it can be seen that, among the three components of non-interest income, a larger share of commission and other non-interest activities leads to a decrease in efficiency for smaller banks, whereas they would gain efficiency improvement when further engaging in trading activities. This finding is consistent with the previous results for large-sized G-SIBs and mediumsized D-SIBs.

148

5 The Impact of Income Diversification on Chinese Banks …

5.4.3 Discussion of the Results Evidence regarding the diversification effect on bank efficiency in China indicates that while diversification yields a discount to the efficiency of the overall Chinese banking sector, the negative effects are mainly concentrated among the G-SIBs and D-SIBS. For N-SIBs, which are relatively small banks, no significant evidence for such an adverse effect is found. The difference in the results among the three groups seems to be associated with bank size and the related issues. As organizations become more complex owing to an increased number of correlated business lines generated from non-interest activities, monitoring becomes more difficult and thus monitoring costs increase (Laeven and Levine 2009). Moreover, bureaucratic problems are more pronounced in large banks, and this can lead to less-efficient operating outcomes. The inefficiencies are further related to the additional overhead costs, inefficient cross-subsidization and moral hazard problems (Klein and Saidenberg 2010). In particular, when large banks—especially systemically important banks—become ever larger, they will be perceived to be TBTF (too big to fail), and regulatory authorities will provide those banks facing serious trouble with rescue packages (Brewer and Jagtiani 2013). This will create a situation in which managers have incentives to accentuate moral hazard and therefore operate such banks in an inefficient manner. The TBTF problem is prevalent in China, and the Chinese authorities would routinely intervene to support large-sized, and invariably state-owned, banks. As a result, large banks in China would be burdened with the moral hazard problem and, as elsewhere, have incentives to expand high-risk but more-profitable projects, causing increased risk and inefficiency of resource allocation (Hellmann et al. 2000). Furthermore, larger banks may find it more difficult to avoid information asymmetry and the associated problems. According to De Jonghe et al. (2015), large banks can obtain diversification benefits only if the information environment and institutional setting allow their stakeholders to exercise proper discipline and when there are no incentives to abuse conflicts of interest that lead to inefficiency. In the Chinese context, information asymmetry is strong in large-sized state-owned banks, but less strong in medium-sized joint-stock banks, and weakest in the smaller-sized city banks and rural banks. Therefore, it will be more difficult for the large-sized banks to monitor the expansion of business lines, and easier for smaller banks within the N-SIBs sub-group to monitor the non-interest activities. Hence, smaller banks are able to achieve efficiency with increased levels of diversification. The structure of banks’ business mode also matters. With regard to the relation between the diversification index and efficiency scores, it has been demonstrated that larger banks (G- and D-SIBs) receive a diversification discount when the diversification index increases. Looking deeper into the effects of component non-interest activities, this result could be driven by a large proportion of fee-based and other non-interest income, which negatively affect efficiency, whereas income from trading activity has a positive effect on efficiency.

5.4 Empirical Estimation

149

It is more revealing to consider the diversification effects from a perspective that combines the banking groups and component non-interest activities. According to the statistics in this study, G-SIBs have the highest levels of incomes from both fee and commissions (11.1%) and other income (1.4%), followed by D-SIBs (8.1% and 0.9%, respectively), with N-SIBs having the lowest level of incomes from these components (4.3% and 0.8%, respectively). However, at the same time, N-SIBs hold the highest proportion of trading income in their total non-interest income and a greater balance between trading and fee-based incomes. It follows that the efficiency discount that the larger banks in China receive could be induced by the fact that they are more likely to diversify towards low-efficiency fee-based activities rather than high-efficiency trading activities, whereas smaller banks are able to maintain more efficient and balanced diversification strategies. Consequently, G-SIBs suffer from high levels of inefficiency, and D-SIBs also experience a reduction in efficiency, albeit to a lesser extent, whereas N-SIBs benefit from higher efficiency. The different effects of diversification on efficiency across bank groups may also be related to the threshold effects of diversification on the level of risk to which banks are exposed. In Chap. 4, we assess the effects of income diversification on risk among Chinese systemically important banks using a threshold model. The results reveal the existence of inverse U-shaped relationships between bank risk and fee-based activity, and between bank risk and trading activity, indicating that there is a threshold point for both of those activities: below the threshold, banks will become less stable, whereas once the banks mature to pass the threshold, they will benefit from increased stability. Compared with fee-based activity, trading activity has a much lower threshold point, where banks can achieve risk reduction with only 0.265% of trading income, rather than 13.899% for fee-based income. In Chap. 4, we also find that the positive effect on banks’ overall risk after passing the threshold point is mainly driven by the trading income rather than fee-based income, whereas other activities continue to bring enhanced risk in both the lower and upper regimes. The fee-based income in the majority of G- and D-SIBs is below the threshold point (with mean values of 10.84 and 7.4%). Then, any increase in fee-based activities by G- and D-SIBs could bring higher risk, in addition to inefficiency. However, the smaller N-SIBs maintain lower levels of both fee-based and other income, whereas the proportion of trading income is greater. For them, this business structure compensates for the diversification discount generated from the high-volatility of fee-based income in the early stages of their development of diversification. From these results, one can expect that the efficiency of Chinese banks would not increase until after the scale of fee-based income for large banks has expanded to pass a certain threshold.

5.5 Conclusion Using a two-step approach, this chapter examines the efficiency implications of Chinese banks’ shift towards a greater share of non-traditional income in their total income. First, efficiency scores of Chinese banks, both cost and profit, are calculated

150

5 The Impact of Income Diversification on Chinese Banks …

via stochastic frontier analysis using the method of within maximum likelihood estimation (WMLE). The analysis is applied to the whole Chinese banking sector as well as to three sub-groups, i.e. global systemically important banks (G-SIBs), domestic systemically important banks (D-SIBs) and other banks (N-SIBs). The results show that in the sample period from 2005 to 2016, the average of both cost and profit efficiency scores first exhibited an increasing tendency from 2005 to 2008 and then began to decline, reaching the lowest point in 2010. Subsequently, the efficiency of Chinese banks improved steadily, to achieve a relatively high point after the global financial crisis. The three sub-groups each followed a similar trend to that of the overall banking sector. In the second step, the investigation employs a dynamic Tobit model to examine the unobserved time-invariant bank heterogeneity. We find that for the overall Chinese banking sector, income diversification has an efficiency-destroying effect. However, the effects vary across the banking groups. For Chinese G-SIBs, diversification has a significant harmful effect on both cost and profit efficiency. For D-SIBs, the effects are similar, but the discount is less. For N-SIBs, diversification has a positive effect on their efficiency level. The differences in empirical results could be explained by the additional overhead cost, inefficient cross-subsidization and moral hazard problems. After decomposing non-interest activities into three components (fee-based, trading and other activities), it is found that the diversification discount is generated from fee-based and other income activities, whereas trading activity can improve banks’ efficiency level. The result is shaped by banks’ internal business structure, as larger banks have an incentive to expand highly volatile and less-effective fee and other non-interest incomes, but smaller banks are more likely to diversify towards less risky and more effective trading activity.

References Abbott M, Wu S, Wang WC (2013) The productivity and performance of Australia’s major banks since deregulation. J Econ Financ 37(1):122–135 Adzobu LD, Agbloyor EK, Aboagye A (2017) The effect of loan portfolio diversification on banks’ risks and return: evidence from an emerging market. Manag Financ Aggelopoulos E, Georgopoulos A (2017) Bank branch efficiency under environmental change: a bootstrap DEA on monthly profit and loss accounting statements of Greek retail branches. Eur J Oper Res 261(3):1170–1188 Aigner D, Lovell CK, Schmidt P (1977) Formulation and estimation of stochastic frontier production function models. J Econom 6:21–37 Akhigbe A, Stevenson BA (2010) Profit efficiency in US BHCs: effects of increasing non-traditional revenue sources. Q Rev Econ Financ 50(2):132–140 Alhassan AL (2015) Income diversification and bank efficiency in an emerging market. Manag Financ 41:1318–1335 Alhassan AL, Tetteh ML (2017) Non-interest income and bank efficiency in Ghana: a two-stage DEA bootstrapping approach. J Afr Bus 18:124–142 Allen F, Gale D (2004) Competition and financial stability. J Money Credit Bank 36(3):453–480 Allen F, Carletti E, Marquez R (2011) Credit market competition and capital regulation. Rev Financ Stud 24(4):983–1018

References

151

Amidu M, Wolfe S (2013) Does bank competition and diversification lead to greater stability? evidence from emerging markets. Rev Dev Financ 3(3):152–166 Ariss RT (2010) On the implications of market power in banking: evidence from developing countries. J Bank Financ 34:765–775 Azzalini A (1985) A class of distributions which includes the normal ones. Scand J Stat 12(2):171– 178 Bakke TE, Gu T (2017) Diversification and cash dynamics. J Financ Econ 123:580–601 Barth JR, Lin C, Ma Y, Seade J, Song FM (2013) Do bank regulation, supervision and monitoring enhance or impede bank efficiency? J Bank Finance 37:2879–2892 Battese GE, Coelli TJ (1988) Prediction of firm-level technical efficiencies with a generalized frontier production function and panel data. J Econom 38:387–399 Bauer PW (1990) Recent developments in the econometric estimation of frontiers. J Econom 46:39– 56 Beccalli E, Frantz P (2009) M&A operations and performance in banking. J Financ Serv Res 36:203–220 Beck T, De Jonghe O, Schepens G (2013) Bank competition and stability: cross-country heterogeneity. J Financ Intermed 22:218–244 Belotti F, Ilardi G (2012) Consistent estimation of the “true” fixed-effects stochastic frontier model. CEIS Working paper No. 23/201 Bencivenga VR, Smith BD (1991) Financial intermediation and endogenous growth. Rev Econ Stud 58:195–209 Berger AN, Humphrey DB (1997) Efficiency of financial institutions: international survey and directions for future research. Eur J Oper Res 98(2):175–212 Berger AN, Mester LJ (1997) Inside the black box: what explains differences in the efficiencies of financial institutions? J Bank Financ 21:895–947 Biener C, Eling M, Wirfs J (2016) The determinants of efficiency and productivity in the Swiss insurance industry. Eur J Oper Res 248(2):703–714 Brewer E, Jagtiani J (2013) How much did banks pay to become too-big-to-fail and to become systemically important? J Financ Serv Res 43:1–35 Busch R, Kick T (2009) Income diversification in the German banking industry. Bundesbank Series 2 Discussion Paper No. 09/2009 Cavallo L, Rossi SP (2002) Do environmental variables affect the performance and technical efficiency of the European banking systems? a parametric analysis using the stochastic frontier approach. Europ J Financ 8(1):123–146 Charnes A, Cooper WW, Rhodes E (1978) Measuring the efficiency of decision making units. Eur J Oper Res 2:429–444 Chen Y, Chen H (2015) The non-interest incomes and operational performance of city commercial banks. J Shanghai Univ Financ Econ 2:51–59 (in Chinese) Chen YY, Schmidt P, Wang HJ (2014) Consistent estimation of the fixed effects stochastic frontier model. J Econom 181:65–76 Cheng MY (2015) Non-interest and bank’s efficiency: evidence from Chinese commercial banks. Mod Econ Sci 37(4):49–126 Chi G-T (2006) Research on the relationship between income structure and income efficiency of China’s commercial banks. J Syst Eng 21:574–590 Cornett MM, McNutt JJ, Tehranian H (2006) Performance changes around bank mergers: revenue enhancements versus cost reductions. J Money Credit Bank 38(4):1013–1050 Curi C, Lozano-Vivas A, Zelenyuk V (2015) Foreign bank diversification and efficiency prior to and during the financial crisis: does one business model fit all? J Bank Financ 61:S22–S35 De Jonghe O, Diepstraten M, Schepens G (2015) Banks’ size, scope and systemic risk: what role for conflicts of interest? J Bank Financ 61(1):S3–S13 Delis M, Tsionas E (2009) The joint estimation of bank-level market power and efficiency. J Bank Financ 33:1842–1850

152

5 The Impact of Income Diversification on Chinese Banks …

Dell’Ariccia G, Igan D, Laeven LU (2012) Credit booms and lending standards: evidence from the subprime mortgage market. J Money Credit Bank 44:367–384 Demirgüç-Kunt A, Huizinga H (2010) Bank activity and funding strategies: the impact on risk and returns. J Financ Econ 98(3):626–650 Deng SE, Elyasiani E (2008) Geographic diversification, bank holding company value, and risk. J Money Credit Bank 40:1217–1238 Doan AT, Lin KL, Doong SC (2017) What drives bank efficiency? the interaction of bank income diversification and ownership. Int Rev Econ Financ 55:203–219 Dong Y, Firth M, Hou W, Yang W (2016) Evaluating the performance of Chinese commercial banks: a comparative analysis of different types of banks. Eur J Oper Res 252:280–295 Dong Y, Hamilton R, Tippett M (2014) Cost efficiency of the Chinese banking sector: a comparison of stochastic frontier analysis and data envelopment analysis. Econ Model 36:298–308 Drake L, Hall MJ, Simper R (2009) Bank modelling methodologies: a comparative non-parametric analysis of efficiency in the Japanese banking sector. J Int Financ Mark Inst Money 19(1):1–15 Drucker S, Puri M (2008) On loan sales, loan contracting, and lending relationships. Rev Financ Stud 22(7):2835–2872 Elsas R, Florysiak D (2015) Dynamic capital structure adjustment and the impact of fractional dependent variables. J Financ Quant Anal 50(5):1105–1133 Elsas R, Hackethal A, Holzhäuser M (2010) The anatomy of bank diversification. J Bank Financ 34(6):1274–1287 Elyasiani E, Wang Y (2012) Bank holding company diversification and production efficiency. Appl Financ Econ 22:1409–1428 Elyasiani E, Staikouras SK, Dontis-Charitos P (2016) Cross-Industry product diversification and contagion in risk and return: the case of bank-insurance and insurance-bank takeovers. J Risk Insur 83:681–718 Feng G, Zhang X (2012) Productivity and efficiency at large and community banks in the US: a Bayesian true random effects stochastic distance frontier analysis. J Bank Financ 36(7):1883– 1895 Fenn P, Vencappa D, Diacon S, Klumpes P, O’Brien C (2008) Market structure and the efficiency of European insurance companies: a stochastic frontier analysis. J Bank Financ 32(1):86–100 Fiorentino E, Karmann A, Koetter M (2006) The cost efficiency of German banks: a comparison of SFA and DEA. Banking and Financial Supervision No. 10/2006 Girardone C, Molyneux P, Gardener EPM (2004) Analysing the determinants of bank efficiency: the case of Italian banks. Appl Econ 36(3):215–227 Gourlay AR, Ravishankar G, Weyman-Jones TG (2006) Non-parametric analysis of efficiency gains from bank mergers in India. Loughborough University Working paper No. 2006–2017 Greene W (2005) Fixed and random effects in stochastic frontier models. J Prod Anal 23(1):7–32 Greene WH (2008) The measurement of productive efficiency and productivity growth. Oxford University Press, Oxford Greene WH (1980) Maximum likelihood estimation of econometric frontier functions. J Econom 13(1):27–56 Gurbuz AO, Yanik S, Ayturk Y (2013) Income diversification and bank performance: evidence from Turkish banking sector. J BRSA Bank Financ Mark 7(1):9–29 Havranek T, Irsova Z, Lesanovska J (2016) Bank efficiency and interest rate pass-through: evidence from Czech loan products. Econ Model 54:153–169 Hellmann TF, Murdock KC, Stiglitz JE (2000) Liberalization, moral hazard in banking, and prudential regulation: are capital requirements enough? Am Econ Rev 90(1):147–165 Hicks JR (1935) Annual survey of economic theory: the theory of monopoly. J Econom Soc 3(1):1– 20 Hjalmarsson L, Veiderpass A (1992) Efficiency and ownership in Swedish electricity retail distribution. J Prod Anal 3:7–23 Hu L (2002) Estimation of a censored dynamic panel data model. Econometrica 70(6):2499–2517 Ibragimov R, Jaffee D, Walden J (2011) Diversification disasters. J Financ Econ 99(2):333–348

References

153

Jiang C, Yao S, Feng G (2013) Bank ownership, privatization, and performance: evidence from a transition country. J Bank Financ 37(9):3364–3372 Kao C, Liu ST (2009) Stochastic data envelopment analysis in measuring the efficiency of Taiwan commercial banks. Eur J Oper Res 196(1):312–322 Klein PG, Saidenberg MR (2010) Organizational structure and the diversification discount: evidence from commercial banking. J Ind Econ 58:127–155 Koetter M, Kolari J, Spierdijk L (2008) Efficient competition? testing the “quiet life” of US banks with adjusted Lerner indices. Federal Reserve Bank of Chicago Working paper No 1094 Köhler M (2014) Does non-interest income make banks more risky? retail-versus investmentoriented banks. Rev Financ Econ 23(4):182–193 Kraft E, Hofler R, Payne J (2006) Privatization, foreign bank entry and bank efficiency in Croatia: a fourier-flexible function stochastic cost frontier analysis. Appl Econ 38(17):2075–2088 Kumbhakar SC, Lovell CK (2003) Stochastic frontier analysis. Cambridge University Press, Cambridge Kumbhakar SC, Wang D (2007) Economic reforms, efficiency and productivity in Chinese banking. J Regul Econ 32:105–129 Kwan SH (2006) The X-efficiency of commercial banks in Hong Kong. J Bank Financ 30:1127–1147 Laeven L, Levine R (2009) Bank governance, regulation and risk taking. J Financ Econ 93:259–275 LaPlante AE (2015) A comprehensive study of bank branch growth potential and growth trends through the development of a unique DEA formulation and a new restricted DEA model. Unpublished Manuscript Lensink R, Meesters A (2014) Institutions and bank performance: a stochastic frontier analysis. Oxford Bull Econ Stat 76:67–92 Lepetit L, Nys E, Rous P, Tarazi A (2008) Bank income structure and risk: an empirical analysis of European banks. J Bank Financ 32:1452–1467 Li T, Zheng X (2008) Semiparametric Bayesian inference for dynamic Tobit panel data models with unobserved heterogeneity. J Appl Econom 23(6):699–728 Lin JR, Chung H, Hsieh MH, Wu S (2012) The determinants of interest margins and their effect on bank diversification: evidence from Asian banks. J Financ Stab 8(2):96–106 Liu C, Ji L (2014) The diversification effects on efficiency: evidence based on Chinese commercial banks. J Stat Decis 23:169–172 (in Chinese) Loudermilk MS (2007) Estimation of fractional dependent variables in dynamic panel data models with an application to firm dividend policy. J Bus Econ Stat 25(4):462–472 Meeusen W, Van den Broeck J (1977) Efficiency estimation from Cobb-Douglas production functions with composed error. Int Econ Rev 435–444 Meng X, Cavoli T, Deng X (2017) Determinants of income diversification: evidence from Chinese banks. Appl Econ 1–18 Mertens A, Urga G (2001) Efficiency, scale and scope economies in the Ukrainian banking sector in 1998. Emerg Mark Rev 2(3):292–308 Meslier C, Morgan DP, Samolyk K, Tarazi A (2014) The benefits of intrastate and interstate geographic diversification in banking. Unpublished manuscript Papke LE, Wooldridge JM (2008) Panel data methods for fractional response variables with an application to test pass rates. J Econom 145:121–133 Pasiouras F, Sifodaskalakis E, Zopounidis C (2007) Estimating and analysing the cost efficiency of Greek cooperative banks: an application of two-stage data envelopment analysis. Int J Financ Serv Manag 5(1):34–51 Pennathur AK, Subrahmanyam V, Vishwasrao S (2012) Income diversification and risk: does ownership matter? an empirical examination of Indian banks. J Bank Financ 36:2203–2215 Quaglia L, Spendzharova A (2017) The conundrum of solving ‘Too Big to Fail’ in the European Union: supranationalization at different speeds. J Common Mark Stud 55(5):1110–1126 Sanya S, Wolfe S (2011) Can banks in emerging economies benefit from revenue diversification? J Financ Serv Res 40(1–2):79–101

154

5 The Impact of Income Diversification on Chinese Banks …

Saunders A, Walter I (1994) Universal banking in the United States: what could we gain? what could we lose?. Oxford University Press, Oxford Schaeck K, Cihak M (2010) Competition, efficiency, and soundness in banking: an industrial organization perspective. In: European banking center discussion paper No. 2010-20S Souza GDS, Gomes EG (2015) Management of agricultural research centers in Brazil: a DEA application using a dynamic GMM approach. Eur J Oper Res 240(3):819–824 Stiroh KJ (2000) How did bank holding companies prosper in the 1990s? J Bank Financ 24(11):1703–1745 Stiroh KJ (2004) Diversification in banking: is noninterest income the answer? J Money Credit Bank 36(5):853–882 Stiroh KJ, Rumble A (2006) The dark side of diversification: the case of US financial holding companies. J Bank Financ 30(8):2131–2161 Sun J, Harimaya K, Yamori N (2013) Regional economic development, strategic investors, and efficiency of Chinese city commercial banks. J Bank Financ 37(5):1602–1611 Tan Y, Floros C, Anchor J (2017) The profitability of Chinese banks: impacts of risk, competition and efficiency. Rev Account Financ 16(1):86–105 Titova Y (2016) Are board characteristics relevant for banking efficiency? evidence from the US. Corp Gov 16(4):655–679 Vives X (2011) Competition policy in banking. Oxf Rev Econ Policy 27(3):479–497 Wei CL, Liu JL (2007) An analysis on diversification of Chinese commercial banks. China Ind Econ 12:10–21 (in Chinese) Wooldridge JM (2005) Simple solutions to the initial conditions problem in dynamic, nonlinear panel data models with unobserved heterogeneity. J Appl Econom 20:39–54 Wu WH (2008) The second financial reform and the development of financial industry in Taiwan. Rev Pac Basin Financ Mark Policies 11(01):75–97 Xia Q, Huang Y (2017) The impact of income structure changes in Chinese commercial banks. J Commer Econ 23(1):171–173 Xu Y (2011) Towards a more accurate measure of foreign bank entry and its impact on domestic banking performance: the case of China. J Bank Financ 35(4):886–901 Yeboah J, Asirifi EK (2016) Mergers and acquisitions on operational cost efficiency of banks in Ghana: a case of Eco bank and access bank. Int J Bus Manag 11:241–249 Zhang J (2003) X-efficiency in China’s banking sector. Financ Res 6:46–57 (in Chinese)

Chapter 6

Conclusion

Abstract This chapter summarizes the main research findings and the implications of the study, and suggests avenues for future research.

6.1 Main Findings Profound changes in the business modes of the global banking industry have transformed the banks’ income structure over the past four decades. As a result, while the interest margin remains the principal source of income for banks, non-interest income has increased its importance in banks’ total revenue. Amid the global trend of income diversification, Chinese banks lately have also become active in pursing business diversification which has raised significance of non-traditional income in their revenue structure. Shifting to non-traditional business to earn fee-based income represents a major challenge for banks. Whether the shift is beneficial has sparked off fresh interest in the literature and a lively debate that centres on the merits and pitfalls of such diversification. The current book contributes to this debate by investigating income diversification in the Chinese banking industry as a case of study particularly for bank diversification in emerging market economies which have been too often overlooked in the existing literature. We consider the diversification effects in three aspects, namely the effect on banks’ profits, risk exposure and efficiency scores. In the study, the overall Chinese banking sector is classified into three sub-groups, namely global systemically important, domestic systemically important, and non-systemically important banks. The grouping is to reflect the complexity of the Chinese banking industry which is fast rising to become the largest one of the world. By providing a comprehensive yet wellstructured study, this book offers to improve our understanding of the desirability of and main diversification effects on banks. In Chap. 2, we introduce the background to the rise and development of income diversification in China. A multitude of factors have acted as the driving forces behind the change. These primarily include regulatory changes, growing completions in the banking environment and unfolding of the financial reforms in China. Consequently, © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0_6

155

156

6 Conclusion

non-traditional and fee-based income has become a substantial part of Chinese banks’ total revenue. The following chapters then move to examine the effects of the income diversification process on Chinese banks. The first of them, i.e. Chap. 3 examines to what extent income diversification would affect the profitability of Chinese banks, which is a first step in analysing the performance of the Chinese banking industry in the age of bank diversification. Employing a dynamic SYS-GMM panel data model to evaluate the performance effects of income diversification, this chapter finds that for the Chinese banking sector as a whole there exists a diversification discount, suggesting that a shift from traditional banking business to mixed business lines negatively affects bank performance. However, structurally, the results are rather diverse. After separating the sample banks into three sub-groups, we find that the largest Chinese banks, China’s global systemically important banks or G-SIBs, can gain positive improvements in their performance through diversification. The next group, the domestic systemically important banks or D-SIBs, shows a non-significant performance response to the shifting to diversified business. The significant under-performer is the group of China’s non-systemically important banks or N-SIBs. The key factor that drives the performance differences lies in the banks’ capability to reap the benefits of diversification through the learning-by-doing process. Other factors include size of the bank, regulatory differences and other factors such as moral hazard. Chapter 4 puts the focus onto the issue of financial stability, and examines to what extent income diversification affects risk exposure of Chinese banks. Previous research indicates a linear relation between income diversification and risk, finding either a negative correlation that suggests banks should diversify, or a positive correlation that indicates banks should remain focused on core business. However, by adopting the first-differenced GMM estimator for the dynamic threshold panel data model, we unearth the evidence showing that in the Chinese case, the relation is not monotonously linear. Rather, the effects of diversification vary with time, sources of non-interest activities and measures of risk. For the whole sample, there exists an inverse U-shaped relation between diversification level and risk. Income diversification will reduce bank risk only after the bank has passed a certain threshold of income diversification. This pattern of the relation seems to be driven mainly by the learn-by-doing effect in relation to the expansion of non-interest activities. After dividing the whole sample into three sub-groups, we find that, for G-SIBs which has a dominant position in the Chinese banking industry, business diversification has a significantly negative effect on the banks’ both idiosyncratic risk and financial distress. However, for D-SIBs and N-SIBs, the relation exhibits an inverse U-shape, where in the early diversification stages the banks incur a discount and become less stable, but they become more stable after achieving a certain threshold level of diversification. It is plausible that these differences reflect the learning by doing effect and others such as different diversification strategies and risk preferences.

6.1 Main Findings

157

Results further reveal that, across different business lines, the diversification effects on bank risk are not uniform. Decomposing the revenue from non-traditional activities further into three sub-classes i.e. fee-based income, income from trading activity and other non-interest income, we find that where there is only a low level of diversification, both trading and other non-interest activities will lead to increased exposure to risk; only once the bank has achieved a certain diversification level will these activities start to provide diversification benefits that will lower bank risk. Activities that generate fee-based income will decrease bank risk for G-SIBs and increase risk for N-SIBs regardless of their diversification levels, while for D-SIBs there is an inverse result, positive for banks with a lower level of diversification and negative for more diversified banks. We also examine the diversification effect across different types of risk. Evidence suggests that for G-SIBs and D-SIBs business diversification can reduce credit risks only when the banks have passed a certain threshold point, while for N-SIBs diversification cannot bring any improvement for credit risk regardless of diversification level. Similarly, the interest rate risk will be reduced only for highly diversified GSIBs and D-SIBs. Finally, diversification will always reduce the liquidity risk, for all three banking groups. However, the reduction of the liquidity risk cannot fully offset the enhancement of the credit and interest rate risks. Therefore, it is necessary for banks to go beyond the threshold in order to obtain the risk-reduction benefits from diversification. This implies that to fully reap the benefits of risk reduction from income diversification, banks need to accumulate sufficient banking human capital and to establish an effective supervision system. The objective of Chap. 5 is to analyse the efficiency implications of Chinese banks’ shift to greater income diversification. To do so, we deploy a two-step approach. First, efficiency scores, both cost and profit, are calculated by stochastic frontier analysis using the method of within maximum likelihood estimation (WMLE). The stochastic frontier analysis is applied respectively to three sub-groups of Chinese banks. The evidence obtained indicates that during the sample period from 2005 to 2016, the average of both cost and profit efficiency scores exhibited an increasing tendency from 2005 to 2008 for the whole sample and then began to decline, reaching its nadir in 2010. Subsequently, the efficiency of Chinese banks improved steadily, to achieve a relatively high point after the global financial crisis. The three sub-groups each followed a similar trend to that of the overall banking sector. In the second step, our investigation employs a dynamic Tobit model to examine the unobserved, time-invariant bank heterogeneity. We find that for the overall Chinese banking sector, income diversification has an efficiency-destroying effect. However, the effects vary across the banking groups. For Chinese G-SIBs, diversification has a significant harmful effect on both cost and profit efficiency. For D-SIBs, the effects are similar, but the discount is less. For N-SIBs, diversification has a positive effect on the efficiency level. The differences in empirical results could be explained by the additional overhead cost, inefficient cross-subsidization and moral hazard problems.

158

6 Conclusion

Further decomposing non-interest activities into three components, i.e. fee-based, trading and other activities, it is found that the diversification discount is generated from fee-based and other income activities, whereas trading activities can improve banks’ efficiency level. The result is shaped by banks’ internal business structure, as larger banks have an incentive to expand fee and other non-interest incomes which are highly volatile and less effective, but smaller banks are more likely to diversify towards less risky and more effective trading activities.

6.2 Implications of the Research Implications flowing from this book have been multiple. The first of them concerns the development prospects of banks’ diversifying into non-traditional activities. Our research shows that expansion of non-interest activities is beneficial for G-SIBs, the most important banks of the Chinese banking system. The performance enhancement can be attributed to the facts that they have accumulated huge assets, resources, technology, and human talent necessary to carry out financial innovations. Chinese D-SIBs are on the borderline, showing some sign of performance improvement, though not strongly significant. With further development of their non-interest business, they can be expected to learn to commend more innovative financial tools and develop better management skills. This implies that they have the potential to grow the diversification further. While N-SIBs show no gains in profitability in the sample period, their efficiency scores are shown to have improved steadily. With this, conducting of non-traditional business in the future can become profitable for them. All these indicates that business diversification by Chinese banks has the room to grow and develop in future. Against this background, it is sensible for managers of G-SIBs to adopt a business expansion strategy that highlights efficiency enhancement because their efficiency scores are low. This can be achieved through using their institutional strength including extensive network of sub-branches to expand fee and commission activity. For small and medium-sized banks, i.e. D-SIBS and N-SIBs the sensible business strategy should focus on providing services through their close ties with customers and gradually develop non-interest financial services to reap the benefits of efficiency enhancement from the process. Structurally, given the fact that income from trading activity has played a particularly positive role in promoting banks’ profitability and efficiency, managers of G-SIBs and D-SIBs should take steps to focus on trading business. For the regulator, our research suggests a structured approach to supervision and regulation over Chinese banks’ conducting of non-traditional business. The existing Chinese regulation over mixed banking business is modelled on the America’s Financial Services Modernization Act of 1999, which is rather restrictive on income diversification. This should be reformed and it is necessary and desirable for the Chinese regulator to relax restrictions on banks’ engagement in non-traditional business. The key area of reform action is to allow an enabling regulatory framework that

6.2 Implications of the Research

159

releases banks from the existing legal constraints on their development of non-interest business. Within this framework, the regulatory priority should be given to oversee the development of potential systemic risk that banks’ shift to non-interest business may cause. The research shows that wide engagement of banks in income diversification would heighten the systemic risk. For one thing, as a result of business diversification, many banks are now doing very similar business. Their product structure thus becomes isomorphic, which makes them vulnerable to common shocks. For another, long-term business expansion consumes large amounts of capital, while banks’ exposure to credit risk, interest risk and shadow banking risk would be on the rise, which reduce the banks to vulnerability further. In addition, relative to the income from traditional business, non-interest income is often instable. Then, with the growth of business diversification, the instability of non-interest income will also grow and the income instability can be transmitted from one bank to other banks. These would amplify the eventuality of systemic risks and the regulator should be wary of this and put the eventuality high in its monitoring radar. This is especially so for some high-leveraged and risky non-interest business requires. On top of close supervision of systemic risk, the book suggests a structural approach to Chinese regulator’s monitoring of financial stress and risk exposure of different banking groups. The inverse U-shaped relation between return and risk in the diversified business implies that some Chinese banks would initially have low performance with heightened risk and only after having passed some threshold would the situation becomes better. This threshold effect needs to be taken into consideration by the regulator in their policy design for supervising the banks, especially over the small and medium sized banks. For the risk file of individual banks, the research has proved that engaging in noninterest businesses will reduce the risk exposure of G-SIBs. But this finding does not relieve G-SIBs from being put under sound supervision and regulation. Rather, considering that G-SIBs’ efficiency scores are lowered by business diversification, the regulator should focus their supervision on the efficiency level of G-SIBs. For N-SIBs, evidence shows that they would see an increase in their risk exposure, so for these banks the regulator’s main concern should be the dynamics of their risk exposure in relation to diversified business. For particular types of risk, it is shown that income diversification can reduce liquidity risk for all three banking groups. But for credit risk and interest risk, while G-SIBs and D-SIBs can manage to reduce their exposure, there is no evidence that this would also be true for N-SIBs. Given this, the regulator should be particularly watchful for the levels of credit and interest risks of N-SIBs. They are relatively small by asset size, but are numerous in numbers and have an extensive customer base. Potential failure of these banks could have far-reaching social repercussions.

160

6 Conclusion

6.3 Limitations and Avenues for Future Research This book attempts to investigate ongoing concerns in understanding the relations between banking diversification in China and its consequences amid the global trend of banks’ shifting to non-traditional businesses. By offering a case study of income diversification of Chinese banks, this study brings closer to a better understanding of the effects of income diversification on profits, risks and efficiency in the Chinese bank industry and hence contributes to the long debate on the desirability and repercussions of banking diversification in recent decades. To advance the knowledge further in the field, it is sound and meaningful that the results and contributions of the book could be considered in the light of its limitations, which also provides the new avenues that could be explored in future studies. The limitations of the current study can be grouped as follows: First, limitations due to the availability problem of raw data. This can be illustrated by the data problems when adopting the threshold dynamic panel estimator based on the first-differenced GMM method. The study applies this method in order to estimate the diversification-risk relation but is constrained by the severe data availability problem. But the methodology requires a balanced data set to satisfy the first difference process, which cannot be satisfied by raw data in China. In the empirical exercises in this book, we collate data of a sample of 40 Chinese commercial banks and the sample period chosen runs from 2005 to 2016. But in this sample period five of the banks did not exist before 2007. This leads the empirical study to reducing the observations by two years to begin in 2007, making the actual sample size relatively small. With the passage of time, we expect the sample period in future can be extended, which dynamic threshold estimation. In that case, an augmented sample size will contribute to the accuracy of test results, and thus provide more accurate estimation. Second, methods for dealing with missing values of data or incomplete data in the regression analysis. In response to the challenge posed by missing values of data or incomplete data, researchers have proposed several methods to estimating the regression model with missing or incomplete data (Abrevaya and Donald 2017). One approach is to deploy the simulated moment of method that imputes the missing values conditional on the other available data (McFadden 1989). Also known as the method of simulated moments, this method is a structural technique that generates simulated data from the economic model, and then matches their moments with those computed from the available data. Alternatively, one may use the indirect inference method (Smith 1993; Gouriéroux et al. 1993). Using an auxiliary model whose parameters are to be estimated from either observed or simulated data, this approach chooses the parameters of the economic model so that these two sets of estimates are as close as possible. Compared to the method of simulated moments, the indirect inference is quite flexible as it allows use of any of the features of sample statistics as a basis for comparison of moments and data. Based on indirect inference, Gouriéroux et al. (2010) propose a general method that can substantially reduce bias related to T is small and fixed while N is large. Indeed, their approach is generic and works

6.3 Limitations and Avenues for Future Research

161

well for any values of N and T (Gouriéroux et al. 2010). This is particularly useful to explore these methods in future research as the Chinese market is less transparent than other mature one, and researchers often can have only relatively short panel data, which may cause biased estimation in the traditional GMM model. Third, other types of diversification. This research has only considered one type, albeit the major one, of banking diversification, i.e. the income diversification. Future research should be extended to consider the development of some other types of diversification, for example geographical diversification, by Chinese banks. Currently, because of data availability, and that financial statements of regional branches are not available, research on other types of diversification is impracticable. When this improves and with increased availability and improved transparency of banking data in China, future research should make a wider coverage of examination of banking diversification. Finally, alternative research strategy. Empirical analysis is not the only way to study diversification effects. In the Chinese banking sector, each bank has its own characteristics including institutional history, development courses, and relations with government and other institutions. These traits will a bearing on banks’ diversification strategies and business performance. As such, future research can be advanced further to adopt a wide range of methods and modelling strategies, including case studies, in order to shed further lights on the effects of business diversification in the Chinese banking industry.

References Abrevaya J, Donald SG (2017) A GMM approach for dealing with missing data on regressors. Rev Econ Stat 99(4):657–662 Gouriéroux C, Monfort A, Renault E (1993) Indirect inference. J Appl Econom 8(S1):S85–S118 Gouriéroux C, Phillips PC, Yu J (2010) Indirect inference for dynamic panel models. J Econom 157(1):68–77 McFadden D (1989) A method of simulated moments for estimation of discrete response models without numerical integration. Econom: J Econom Soc 995–1026 Smith AA Jr (1993) Estimating nonlinear time-series models using simulated vector autoregressions. J Appl Econom 8(S1):S63–S84

Appendix

See Tables A1, A2, A3 and A4. Table A1 Income diversification and banks’ performance for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 ROA ROA (t − 1)

0.452*** (0.017)

HHI_Dummy_GSIBs

0.005**

HHI_Dummy_DSIBs

0.002

HHI_Dummy_NSIBs

−0.004***

(0.002) (0.001) (0.001) NIM

0.184***

LTA

−0.004

NON

0.009

(0.012) (0.001) (0.031) Constant

0.256*** (0.053) (continued)

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2020 Z. Qu, Income Diversification in the Chinese Banking Industry: Challenges and Opportunities, https://doi.org/10.1007/978-981-15-5890-0

163

164

Appendix

Table A1 (continued) ROA F-test

0.000

Sargan test

0.390

AR(2)

0.123

Observations

408

This table reports the two-step SYS-GMM dynamic panel estimation results with robust errors. Our dependent variables is return on assets (ROA). ROA (t − 1) refers to the lagged dependent variables by one period. HHI_Dummy_GSIBs, HHI_Dummy_DSIBs and HHI_Dummy_NSIBs indicate three interaction terms by using the Herfindahl-Hirschman index multiply three dummy variables to catalogue G-SIBs, D-SIBs and N-SIBs. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively Table A2 Income diversification and banks’ performance for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 Fee and commissions

Trading activities

Other activities

0.442***

0.336***

0.709***

(0.016)

(0.073)

(0.028)

ROA ROA (t − 1) COM_Dummy_GSIBs

0.012** (0.006)

COM_Dummy_DSIBs

−0.011**

COM_Dummy_NSIBs

−0.020***

(0.006) (0.003) TRAD_Dummy_GSIBs

0.365*** (0.136)

TRAD_Dummy_DSIBs

0.052*

TRAD_Dummy_NSIBs

0.036**

(0.030) (0.016) OTH_Dummy_GSIBs

−0.023 (0.019)

OTH_Dummy_DSIBs

−0.064***

OTH_Dummy_NSIBs

−0.035***

(0.013) (0.007) (continued)

Appendix

165

Table A2 (continued) Fee and commissions

Trading activities

Other activities

NIM

0.165***

0.304***

−0.043***

(0.013)

(0.063)

(0.015)

LTA

−0.005***

−0.003

0.001

(0.001)

(0.003)

(0.001)

0.096**

−0.069

0.447***

(0.042)

(0.093)

(0.034)

Constant

0.342***

−0.005

0.050

(0.041)

(0.138)

(0.044)

F-test

0.000

0.000

0.000

Sargan test

0.450

0.992

1.000

NON

AR(2)

0.114

0.301

0.875

Observations

408

399

406

This table reports the two-step SYS-GMM dynamic panel estimation results with robust errors. Our dependent variables is return on assets (ROA). ROA (t − 1) refers to the lagged dependent variables by one period. COM_Dummy_GSIBs, COM_Dummy_DSIBs, COM_Dummy_NSIBs, TRAD_Dummy_GSIBs, TRAD_Dummy_DSIBs, TRAD_Dummy_NSIBs, OTH_Dummy_GSIBs, OTH_Dummy_DSIBs and OTH_Dummy_NSIBs indicate nine interaction terms by using the fee and commissions, trading income and other income multiply three dummy variables to catalogue G-SIBs, D-SIBs and N-SIBs. NIM indicates total interest income/total interest expenses. LTA is loans/total assets, and NON is the non-interest expenses/total assets. The Sargan test checks the null hypothesis, i.e., that the instruments used are not correlated with the residuals. AR (2) denotes the Arellano-Bond test for the 2nd-order autocorrelation in first differences. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively Table A3 Income diversification and banks’ efficiency for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 Cost efficiency Cost_Efficiency (t − 1)

0.689***

HHI_Dummy_GSIBs

−0.191**

HHI_Dummy_DSIBs

−0.313***

(0.164) (0.086) (0.107) HHI_Dummy_NSIBs

0.001

Lerner

−0.472***

(0.093) (0.078) (continued)

166

Appendix

Table A3 (continued) Cost efficiency NIM

3.880***

ETA

0.011

(0.876) (0.356) CIR

−0.139***

Constant

0.455***

Log likelihood

12.464

LR test (p-value)

0.000

Observations

402

(0.049) (0.175)

This table reports the results from Dynamic Tobit estimation. Our dependent variables are cost efficiency, which are estimated by stochastic frontier analysis. Cost_Efficiency (t − 1) refers to the lagged dependent variables by one period. HHI_Dummy_GSIBs, HHI_Dummy_DSIBs and HHI_Dummy_NSIBs indicate three interaction terms by using the Herfindahl-Hirschman index multiply three dummy variables to catalogue G-SIBs, D-SIBs and N-SIBs. Lerner indicates the Lerner index, NIM indicates net interest revenue over total earning assets, ETA refers to equity/total assets, CIR indicates the cost-income ratio. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively

Appendix

167

Table A4 Income diversification and banks’ efficiency for G-SIBs, D-SIBs and N-SIBs by adopting dummy variables, 2005–2016 Fee and commissions

Trading activities

Other activities

Cost efficiency Cost_Efficiency (t − 1) COM_Dummy_GSIBs

0.454***

0.695***

0.498***

(0.042)

(0.148)

(0.054)

−0.562*** (0.107)

COM_Dummy_DSIBs

−0.332*** (0.108)

COM_Dummy_NSIBs

−0.587*** (0.105)

TRAD_Dummy_GSIBs

0.074 (3.265)

TRAD_Dummy_DSIBs

1.377*** (0.468)

TRAD_Dummy_NSIBs

0.558** (0.234) −0.281**

OTH_Dummy_GSIBs

(0.119) −0.063

OTH_Dummy_DSIBs

(0.903) OTH_Dummy_NSIBs

0.466** (0.205)

Lerner NIM ETA CIR Constant

−0.294***

−0.331***

−0.297***

(0.043)

(0.057)

(0.051)

1.938**

3.751***

3.254***

(0.597)

(0.725)

(0.670)

0.896***

0.762**

0.564**

(0.202)

(0.349)

(0.235)

−0.075*

−0.020

−0.011

(0.045)

(0.055)

(0.049)

0.529***

0.245*

0.424***

(0.051)

(0.146)

(0.060)

Log likelihood

589.446

338.381

561.330

LR test (p-value)

0.000

0.034

0.000

Observations

402

393

391

Dynamic Tobit estimation. Our dependent variables are cost efficiency, which are estimated by stochastic frontier analysis. Cost_Efficiency (t − 1) refers to the lagged dependent variables by one period. COM_Dummy_GSIBs, COM_Dummy_DSIBs, COM_Dummy_NSIBs, TRAD_Dummy_GSIBs, TRAD_Dummy_DSIBs, TRAD_Dummy_NSIBs, OTH_Dummy_GSIBs, OTH_Dummy_DSIBs and OTH_Dummy_NSIBs indicate nine interaction terms by using the fee and commissions, trading income and other income multiply three dummy variables to catalogue G-SIBs, D-SIBs and N-SIBs. Lerner indicates the Lerner index, NIM indicates net interest revenue over total earning assets, ETA refers to equity/total assets, CIR indicates the cost-income ratio. Figures in brackets present the standard error. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively